Committed before the repeats finish. Unit is (change, repeat); paired bootstrap
over changes and exact McNemar from openultrasast.stats; failures stay in the
denominator. Dry run on the existing single runs reproduces 6/12 and 3/12.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
3 of 12 with the fixed answer check against the pilot hunter's 6 of 12, one
run each, not significant. The intended zero-cost replay was not a replay:
answers at temperature 0 differ between runs and the cache did not serve the
second run. Programs must be compared over repeated runs, paired by change.
About $1.10 for both runs.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
0 of 12 positives; 12.6 false alarms per 100 ordinary pushes (2.1 on changed
lines only). The misses come from what the rules and facts express (access
control lives in obligations, which do not convert; sources cover request
objects only), not from the engine. Comparison arm only: a Semgrep miss is
never evidence against a hunter finding.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Engine LGPL-2.1, shelling out carries no obligation; the registry rules are
under the Semgrep Rules License since 2024-12-13 and cannot ship, only our
own converted rules can. Measured 4.7 s startup and 5 to 10 ms per file;
converted regex and taint rules fire for Python, PHP and JavaScript; C
propagation and Groovy do not; one sanitizer semantic we refuse on purpose
is accepted by Semgrep. Recommendation: a Semgrep arm on the valid pilot
first. Recorded in the hunter-learning-loop research note and breadth.toml.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
610 pair files, 570 labelled functions. lizard matches the parser engine on
C, Python and Java but loses JavaScript names and TypeScript end lines:
keep ours, take its cyclomatic complexity as a feature candidate.
tree-sitter-language-pack gives identical spans, is faster, replaces six
grammar wheels with one and adds Go, Ruby, PHP and Perl: adopt as grammar
packaging. Also found: .c files should try the c grammar before cpp.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
src/openultrasast/source.py replaces the copies in semantic/functions.py,
plane/tasks/repo_facts.py, learn/features.py, learn/excerpt.py and
plane/tasks/verify.py. Two engines share one sentinel (<module>), one
DECLARATION/KEYWORDS/RESERVED set and one alias map: the parser engine
(ranges, enclosing, bounds, call_sites, callers_of, facts_for) for the
scorer, the overlay and the hunter's annotations; the line-pattern engine
(declared_functions, enclosing_at, function_span, function_bounds, the brace
matcher) that labels, features, roles, excerpts and plane facts were built
on. learn no longer imports plane.tasks. Frozen records that store the old
sentinel are mapped on read (module_function).
Gate: labels rebuilt before and after are identical, 7,921 rows, after
sentinel normalisation (benchmarks/measurements/2026-10-13-source-module-
label-diff/record.json). 490 targeted tests pass; 3,312 collect. Net: 657
lines deleted, 212 added. benchmarks/unseen keeps <global> for its own
frozen manifests.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Off every active line (PATHFINDER-2026-10-13/03-unified-proposal.md, cuts).
Deleted: the four hunter tools no caller enables (read_definition,
entry_points, flows, obligations) with their schemas and dispatch, the dead
mapping branch of find_refs; 20 functions whose only caller was one test
(mapping task records, compute_ranking_metrics, build_quick_hunter_prompt,
validate_evidence_transition, should_fuse, harness processors,
rank_targets_with_model, model/report slice rendering, taxonomy
hierarchical/related, config_value.verdict, taint.verdict, roles.load_roles,
features.language_of and build_for_scan, folds.leave_one_*,
features.recorded_case, search BatchExecutor and the three __main__ shims,
delivery.put_input); the receiver delivery mode of the plane. push_inputs.py
moved to benchmarks/push. memory.purge stays: the live-S3 test cleanup uses
it. Net: 2,730 lines deleted, 207 added. 473 targeted tests passed; 3,282
tests collect. breadth.toml records the removals.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Off every active line (PATHFINDER-2026-10-13/03-unified-proposal.md, cuts).
Deleted: src/openultrasast/regress/ and the deep-mode REGRESS branch (deep
now records a regress_retired degradation and equals standard; proof is the
search-with-proof arm's job), the sandbox docker runner and probe that only
regress used, --fail-on worth-fixing, RegressConfig, the two ledger helpers
only regress called; skills.py and index.py with the hunter pool's skill
routing; model/audit.py; cpg/queries/vocabulary.sc with
benchmarks/push/ontology_gap.py. Docs and the scan skill describe the
retirement. Net: 3,895 lines deleted, 86 added. Targeted tests: 180 passed,
28 skipped; 3,331 tests collect.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Sweep over the pilot's hunter records: most 'vulnerable' suspicions on ordinary
pushes sit far from the diff. Scope to the change, then confidence, then the
config_secrets family; 5/12 recall kept at 8.8 false alarms per 100 (in-sample).
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Opt-in filter on the classifier verdict for hunter-sourced defects; the default
keeps everything. Each combined record notes the kept verdicts and the counts.
Live check on the pilot: (a)+(b) filtered to 'vulnerable' gives recall 5/12 and
54.4 false alarms per 100 ordinary pushes, matching the scratch numbers.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Arm (b) hunter over all 252 changes: recall 6/12 raw, 5/12 with the classifier
filter; false alarms 92.5 and 54.4 per 100 ordinary pushes. Arm (a) push hook at
163/252: recall 0/12, 0.7 false alarms per 100. Per-positive table, classifier
stats, the one hook false alarm, and the disk fix for the replay output.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Live run: every classification failed with ValueError because Caller.ask
reads client.usage as a dict (the learn CLI always hands it a MeteredClient)
while a bare DeepSeekChatClient keeps usage as a list of rows. build_caller
now wraps a client that carries a usage list in MeteredClient; scripted and
already-metered clients pass through unchanged.
Regression test runs classify_family through the real DeepSeekChatClient
wrapper with a stub transport (the module fixture that stubs classify is
restored for it). The error field stays the exception type only, since a
message can carry a model key.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The tool hunter's family tag is the model's free text (a live run returned
family:sql-parser and family:readonly-enforcement), so every arm (b) defect
scored family "unknown" and score.catches() could never credit the hunter
even when it landed on the labelled file.
hunter_arm now classifies each finding into the closed taxonomy vocabulary
with the decision-engine classifier (learn.program.classify with
families=FAMILY_IDS minus unknown): a numbered, clamped excerpt around the
finding sets valid_lines, the canonical ANSWER_FORMAT is reused, an off-list
family parses to "none" and lands as "unknown". The benchmark label is never
a fallback. The defect keeps the raw hunter family, title and rationale plus
the classifier's verdict, confidence, rationale, cited lines and parse
failures for audit; the record counts classifier calls and families.
--no-classify keeps the old heuristic path.
14 new offline tests, 67 green with the benchmark tests.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
benchmarks/valid/hunter_arm.py runs the shipped deep-mode tool hunter
(tool_hunter.run_tool_hunter via resolve_hunter_client) over each frozen
change off-cluster: detached worktree at head with git hooks disabled,
hotspots scoped to the changed files, instrument rule on every hotspot
(bytes read printed, empty or unreadable fails the change loudly), model id
from the config resolver, family from the taxonomy, never from the label.
Records land in the exact shape score.py reads (push.admission.defects with
locations), resumable per change.
`combine` joins arm (a) replay records with arm (b) hunter records per change
id, dedupes defects by (path, line, family), keeps (a)'s instrument, records
which arms were present, and writes a file score.py loads directly.
9 offline tests (stub hunter, real scorer), 53 green with the benchmark
tests. No network, no key handling, no live run.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
First live run of the valid detection benchmark. Labels come from the fix
site (fix^1), never SZZ.
- 32 held-out Python/JS advisories sourced from the public pool-p1 set,
CWE + text from OSV. 133 candidate locations -> 125 after dropping
tests/minified.
- 125 DeepSeek verification calls: 31 locations passed all three checks
(vulnerable pre-fix, fix addresses it, reachable in-repo) across 12 repos.
- Maintainer spot-check accepted all 12.
- 20 ordinary negatives per positive change from the same repos, 240 total.
- Held-out guard: 0 dropped against the training-only used set.
- Frozen manifest at benchmarks/valid/pilot/ (252 changes), provenance and
NOTES under benchmarks/measurements/2026-10-12-valid-benchmark-pilot/.
Three fixes found by the live run:
- catalog.py: skip a site the fix renamed or removed instead of aborting.
- assemble.py: negative ratio counts positive changes, not accepted
functions (malla would have demanded 260 negatives). Test updated.
- held-out guard: run against training sources only; build_used() also
scans measurements/populations, which flagged the 12 repos via my own
SZZ-pool analysis outputs (circular).
Replay and metrics not yet run; see NOTES.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Builds the benchmark from .kiro/specs/valid-detection-benchmark/brief.md:
sound labels from the fix (no SZZ), app-shaped verified positives,
ordinary negatives from the same repos, whole tool on unseen repos,
standard metrics.
Spec: requirements.md, design.md, tasks.md. Reuses the existing sourcing,
fix-site labelling, negative drawing, replay, and Wilson scoring; drops
the SZZ layer.
New offline modules under benchmarks/valid/:
- catalog.py positives labelled at the fix site (never introducing/last_touch)
- verify.py one LLM call per case (vulnerable? fix addresses it? reachable
in-repo?) + human spot-check; this is the app-shaped filter.
Honors the instrument rule: bytes read per function, reject on
an empty read.
- assemble.py >=20x ordinary negatives + a frozen, deterministic manifest with
the held-out guard
- metrics.py precision (PPV), false alarms per 100 pushes, exploit-verified
rate, and the plain headline, layered on score.py
36 new offline tests, 495 regression tests green. No network, no harvest,
no model calls in tests. Group 5 (the live pilot) is not started.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Roadmap "Where we are" table said Baseline (G0.2) = "Not yet, run on kube" -- stale.
Record 07 (G0 exit) was completed right after on the docker lane (e901640, 1300/1300,
catch 3.3% / FP 3.1%, regex/quick-rule floor, engine_covered=0). G0 is exited; the kube
`ax` lane was never needed for the baseline. Reframed the readiness plan: steps 1-4 done,
the `ax` lane moves to G1 (engine arm at scale, gated by spec tasks 4.4/5.4).
Also fixed the ax binary in the coordinator commands: use $HOME/go/bin/ax, NOT
$HOME/.local/ax-v0.3.1/ax (the latter hits the AX-lane dispatch blocker recorded in the
unseen spec tasks.md). README.md + benchmarks/unseen/README.md.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
test_no_pair_repeats_a_repository_the_benchmarks_already_name failed: advisory-fix
catalog repos also appeared elsewhere under benchmarks/. Root cause split two ways:
1. Genuine leak: 16 catalog repos are also subjects in benchmarks/search/private/
pilot.json (a real private pilot population). Removed those 16 pairs from the
three catalogs and their verification.json rows (totals recomputed, catalog ==
verification preserved).
2. Spurious: the overlap scan greped ~49k vendored files under
benchmarks/search/private/runs/*/checkout/node_modules/, so any catalog repo a
vendored dependency's README names collided. Added a narrow vendored-tree skip
(node_modules/vendor) to the scan rather than deleting labelled pairs over a
dependency mention.
Implemented from Codex's first pass (gpt-6.1-sol), then gated and corrected: Codex
removed 20 pairs; 4 of those (backstage/backstage, koajs/koa, beaugunderson/
ip-address, cakephp/cakephp) were driven only by vendored/non-population mentions
and are kept. Also excluded advisory-fixes-spread.json (catalog-derived report)
and research-2026-09-26.md (selection write-up) as non-population.
tests/test_advisory_fix_pairs.py + tests/test_learn_labels.py: 53 passed.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
taint.request_params emits guards/fixedOrigin/originAnchors/quotedSanitizers/
fieldParameterSources, which taint.sc reads from per-request JSON in its BATCH
branch but does NOT declare on `@main def exec` and does NOT consume in its
single-request branch (it defaults them). The single-query path passed them as
`--param guards=...` (disposable) or named args (session), so joern errored
`Unknown arguments` and every cpg.run("taint", ...) failed -- a silent zero of
the shape AGENTS.md warns about. scan_region->taint_verdict->_usable_flows and
the pair-path verdict()/verdicts() were dead; the push/scan batch route was
unaffected (it carries the extras as JSON the script parses itself).
Fix: parse the script's `@main def exec(...)` signature once (_declared_exec_params
+ memoised _declared_params) and, in query(), filter params to the declared set
before building either route. Faithful, not lossy -- the single-request branch
could never use those 5 fields. No-op for config/dominance (their request_params
emit only declared params; verified). Only taint sheds the 5 extras.
Verified end-to-end on banks (patched backend in the replay image): single-query
path now runs; baseline 0 findings, annotated search 2 path/model_entailed
findings (prompt->write_text). Regression test asserts no undeclared param
reaches the command line and the query parses, with a fake runner (no Joern).
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Same CPG, only the search changes. Baseline (shipped path spec) resolves 0
prompt->write_text flows on banks; injecting the two annotations an LLM steer
would supply (parameter_sources=True + write_text/read_text path sinks) makes the
engine resolve prompt -> path-join -> prompt_file.write_text(prompt.raw). First
measured instance of "#2 guided by #1" closing on an unseen target; fix needed
exactly the param-source + file-write-sink annotations.
Files the instrument bug the first MVP hit: single-query cpg.run("taint",
request_params(...)) is dead -- request_params emits 5 keys (guards/fixedOrigin/
originAnchors/quotedSanitizers/fieldParameterSources) absent from taint.sc's exec
signature; both single-query sub-routes reject them (silent None). Batch path
(run_batch->requestsFile JSON) unaffected; push/scan route uses batch.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Ran ousast scan --mode deep on the 5 app-shaped + sound pool targets (the fair change-scoped
eval set from the shape classification). Detection 3/5: graphiti (hunter, Cypher injection),
mcp-remote (hunter, url-scheme injection), dcmcp (ENGINE overlay -- 'outbound.request reached
from parameter' on readFileFromUrl; parameter-as-source fired because an entry point is present).
2/5 missed: malla (DOM/template XSS) and dms (Jinja SSTI), both in families families.toml
documents as limited.
This validates the strategic pivot: the near-zero catch was substantially a TARGET MISMATCH --
on app-shaped repos the engine itself detects (0 -> 3/5). But all hits stop at suspicion/
static_corroboration; release-grade BLOCK still needs proof-arm elevation + family coverage +
the DSPy hint loop. Cost ~$0.007/repo. So bet #1 (fix eval target) is necessary and material
but not sufficient.
Next: recompute a floor on an app-shaped sound subset / real app corpus (VAmPI, NodeGoat,
DjanGoat, vibe-py); measure proof-arm suspicion->proven lift.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Named the core problem with the maintainer: the tool spots dangerous ops but cannot prove
attacker-reachability -- the trust boundary lives outside the changed library code. Chosen
strategic bet (of three): FIX THE EVALUATION TARGET FIRST.
LLM shape classification of the 49 public pool vulns ($0.033): reachability_in_repo only
24% (12/49); 76% are pure libraries. A legitimate change-scoped catch needs sound label AND
reachability in-repo -> just 5/49 (~10%): graphiti, desktopcommandermcp, malla,
document-merge-service, mcp-remote. Shape predicted the hunter outcome (graphiti app-shaped ->
caught; bedrock library -> missed). So record-07's 3/92 is scored against a ~90%-illegitimate
denominator -- the near-zero catch is largely a target mismatch.
Next: deep scan on the 5 fair targets for a per-tier catch rate; then an app-shaped eval set
from known-vulnerable apps. Artifacts: 2026-10-12-pool-p1-shape-classification/ (+ earlier
label-audit, hunter-sound-targets). Memory: evaluation-target-mismatch.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
ousast scan --mode deep (DeepSeek deepseek-flash, model layer + tool_hunter, 60-call bound) on
bedrock + graphiti, the two sound pilot targets the static engine missed:
- graphiti: HIT -- falkordb_driver.py:207, tag family:cypher-injection; rationale traces
build_fulltext_query concatenating user input into a Cypher string at line 278 (inside the
labelled function, location-matches). The static engine produced zero findings here.
- bedrock: miss -- 12 suspicions on adjacent concerns (network-PUBLIC, download_file), never
named the install_packages -> pip command injection.
Cost is a non-issue: ~$0.007/repo, 64 model calls for a whole-repo deep scan. BUT everything
stopped at suspicion rung -- the proof arm did not elevate to proven; release-grade BLOCK needs
proven findings.
Direction (maintainer): prompt evolution is a core lever -- a DSPy-based loop (GEPA/MIPRO,
pluggable; evolving field) whose objective is to learn what hints/annotations steer the model
to the real sink, sourced from the framework knowledge already shipped (no new user
annotation), with LLM-proposed boundaries graduating into the facts.
Next: measure proof-arm lift on graphiti; stand up the DSPy hint loop + re-run bedrock; widen
the hunter to the 23 sound targets for a catch rate; then a change-scoped + proven run.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Widened the banks label-mislocation finding across all 49 public pool vulns via DeepSeek
(advisory + head code + pool label -> sound/unsound), $0.038 total. Result: 23/49 (47%)
sound; 53% unsound (25 pre-exist in base, 16 sink outside the changed lines) -- a change-scoped
hook correctly stays quiet there. Systematic by family: injection 10/13 sound, but path 1/7,
output_encoding 1/8, config_secrets 1/4.
Implications: record-07's floor (catch 3/92) is scored against a corpus ~half unfair to a delta
hook; the honest catchable denominator is the sound subset (caveat added to record 07 NOTES).
Fair pilot re-score: only bedrock + graphiti (injection) of the 5 are sound and the engine
missed both = the real detection gaps for the LLM-guided hunter.
Also: GEPA-style reflective-retry (future spec) + GLM cyber model (future spec) noted; DeepSeek
is the working default. Artifact: benchmarks/measurements/2026-10-12-pool-p1-label-audit/.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The real bottleneck was not enumeration or extraction: canonicalize_used
resolves the entire exclusion set (measured: 4,499 names) via serial
GitHub get_repo on every launch, before any harvesting -- ~30 min of
network-latency-bound work that also burns ~4,499 of GitHub's 5,000/hr
budget before real work starts (hence the low CPU and the "never
finishes"). The earlier resolution-cache only banked via
BankedGitHub.resolve, but canonicalize_used calls eligibility.resolve ->
get_repo directly, so it was never cached and re-ran every time.
Cache get_repo at the real choke point. CorpusBank gains cache_repo /
get_cached_repo storing one object per normalized name at
<prefix>/repos/<sha256(name)>.json -- NOT seed-scoped, since repo
metadata is seed-independent, so every seed/run shares it. 404s are
cached as None too (a found-but-None hit) so known-missing repos are not
re-fetched. BankedGitHub.get_repo now checks the local journal, then the
bucket, and only then GitHub (banking the result). canonicalize_used,
advisory_index and candidate resolution all funnel through get_repo, so
all of them are covered.
Effect: the first run pays the lookups once and banks them; every later
launch reads them from the bucket (no GitHub, no rate-limit spend), so
the rate budget goes to actual candidate harvesting and restarts are
cheap. Proven live against ousast-corpus (repos/ fills 1:1 with
resolutions). 206 unseen tests + new repo-cache test pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The decision bank made extraction durable, but every restart still
re-paid the ~26-minute advisory enumeration + per-advisory resolution,
since those lived only in the local journal. Mirror them into the bucket
so a restart reads them back and goes essentially straight to any
remaining extraction.
bank.py: bank_resolution/resolved at <prefix>/<seed>/resolved/<sha256(ghsa)>.json
(streamed one-by-one as each advisory is decided), cache_advisories/
cached_advisories at <prefix>/<seed>/advisories/list.json, and
bank_repository/repositories (caches the advisory_index repo-resolution
pass, preserving renames and 404s). All gated by RESOLUTION_CONTRACT="v1"
(extract.py) -- bumped only when enumeration/eligibility/source.candidate
output changes; recent fetch/parse/pool changes do not bump it.
draft.py run(): with --bank, advisories load from the cache (else
enumerate then cache); eligibility decisions load from the bucket
(contract-matched) to skip re-resolution; new resolutions stream to the
bucket before the local checkpoint write. Repository lookups reuse the
repositories/ cache. No --bank => local-journal behavior unchanged.
Net: advisories + resolutions + extraction all resume from ousast-corpus;
restarts no longer re-enumerate or re-resolve. Fresh and resumed runs
produce byte-identical manifests at 1 and 4 workers. Bank methods proven
live against ousast-corpus. 356 tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The draw lost all progress on any restart because the local journal
invalidates on a code change and is the only cross-run state. Make the
bucket the source of truth: stream every per-repo accept/reject to
ousast-corpus as it is decided, keyed by sha256(repository+"\n"+fix) so
writes are idempotent (content dedup at rest), and resume by listing the
bucket and skipping already-decided repos. No supervisor/watchdog needed
-- a crash or a code change just re-runs and resumes.
New benchmarks/unseen/bank.py: corpus_settings (CORPUS_S3_* from .env,
missing key named, value never printed), CorpusObjects (lazy boto3,
path-style + SigV4 verified live against ousast-corpus @ 10.0.0.54:9001),
CorpusBank.bank/decided with one-retry put/get, split_decided. Decisions
live at <prefix>/<seed>/decisions/<key>.json with repository, fix,
accept/reject, reason, entry, exclusions, counts, contract, seed, ts.
draft.py: --bank [prefix] opens the bank on the coordinator (never shipped
to fork workers); run() resumes from bank.decided() before extracting,
reusing accepted entries across contracts (deterministic from repo+commit)
and rejections only when contract == EXTRACTION_CONTRACT ("v1", in
extract.py); collect() streams each decision one-by-one. Seed-order
acceptance, fork/used-set dedup, and duplicate-advisory reuse preserved.
Without --bank, local-journal behavior is unchanged.
CorpusBank proven live end-to-end (bank accept+reject, idempotent re-bank,
decided/split roundtrip). 325 tests pass offline, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The harvester cloned --filter=blob:none --no-checkout, then blame/log/diff
triggered the promisor to fetch each needed blob ON DEMAND -- hundreds of
round-trips per repo, the dominant cost that made the draw unable to
converge. The blobless clone was a premature optimization that backfired.
Drop --filter=blob:none from the clone and the two commit fetches. A
full-objects --no-checkout clone brings the whole commit graph AND every
blob in one pack, so blame/log/diff read locally; --no-checkout still
skips the working tree so disk stays bounded (plus the existing 500 MiB
clone_size_cap and the disk guard). With no promisor remote, a genuinely
missing object now errors loudly instead of silently round-tripping --
the instrument guarantee that the bulk fetch was complete.
Extraction semantics are unchanged (same SZZ introducing-commit positive,
same stratified ordinary negatives, byte-identical entries); only the
object-transfer strategy changes. Test inverted to assert no blob:none
filter on the clone; README updated. 145 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Git subprocess calls were deadline-bounded (Git.command_timeout uses the
remaining per-repo budget), but a pure-Python parse between git calls
(repository_changes -> history/blame/diff on a giant file) had no
deadline: when it hung, no further git call fired the git deadline, so
the worker stalled indefinitely (observed >720s with zero git
processes). An external watchdog was papering over it at ~20 min/stall.
Add extract.process_deadline(seconds): a SIGALRM/setitimer wall-clock
alarm installed on the worker's main thread that raises
ExtractionTimeout("wallclock_timeout"), with no-op guards (None/<=0, off
the main thread, or no SIGALRM support) and strict timer-cancel +
handler-restore in finally. Wrap the whole extract_repository body in it
so the timeout propagates to the existing `except source.Rejected` and
journals as a clean rejection, not a worker failure.
Budget = --extract-deadline (git, 240s) + --parse-deadline (new, 120s)
= 360s total ceiling, so git-bound repos keep their full budget (yield
unchanged) while a hung parse is rejected well before any external
watchdog. Record parse_deadline_seconds in the report.
295 unseen tests pass (incl. timeout + handler-restore + no-op cases),
ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The draw hardcoded 300 repos / 3 slices. At the measured lazy-fetch rate
the full 300 projects ~2.5 days; a 100-repo single slice (slice 1, all
the G0 baseline needs) is ~20h. Add --pool-size (default 300) and
--slices (default 3), validated (>=1, slices<=pool_size, evenly
divisible). Generalize assign_slices' stratified allocation from a
literal 3 to `slices`, the run() break / insufficient check / report
(changes = pool_size*13) to pool_size, and record both in the journal
inputs. Remove bisect's three-slice assumption; freeze reads the size
from the draft. Defaults (300/3) reproduce prior output exactly.
With --pool-size 100 --slices 1 the pool is the first 100 accepted repos
in seed order, all in slice 1 (deterministic, byte-identical across
worker counts). 280 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Git.materialize enumerated every object reachable from HEAD and tried to
`git fetch --no-filter` them by SHA. Git rejects that for real repos
("did not send all necessary objects" / "bad revision"), exiting 1 ->
InstrumentFailure(git_exit_1) that failed the whole repository. Measured
in a live draw: 326/670 attempts, ~49% of all failures, which froze the
accepted pool at 44/300. Reproduced on pallets/flask (26221 reachable,
9287 missing, fetch exit 1).
Git cannot batch-fetch arbitrary tree/blob objects by SHA, so remove
materialize and restore the original lazy path: the blob:none clone keeps
the commit graph and the introducing/blame/ordinary-history reads fetch
their blobs on demand through Git's promisor. The per-fix tolerant fetch
loop, the process-pool parallelism, the 25-with-zero-entries guard, the
per-repo deadline and resume are unchanged. Extraction content is
byte-identical (only the prefetch is gone).
217 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Two defects stalled the draw at 40/300:
1. The systemic guard used an absolute `instrument_failures > 200` cap.
At ~10% extraction yield, reaching 300 entries legitimately produces
far more than 200 per-candidate failures, and the counter is rebuilt
from the journal on resume -- so once past 200 the guard tripped
immediately on every restart (permanent stall; the supervisor's
no-progress guard then stopped). Drop the absolute cap; keep only
`>=25 failures with zero entries`, which still catches a genuinely
broken instrument (it yields no entries) without aborting a working
low-yield draw. Bounded anyway by the population / max-hours.
2. repository_changes fetched all fixes in one `git fetch origin *fixes`,
so a single unfetchable secondary fix ref failed the whole repo
(extract_git_exit_1 was 166 of ~200 extract_ failures -- the dominant
loss). Fetch the extra advisory fixes per-fix, best-effort, skipping a
git_exit_ failure; candidate.fix stays strict (already fetched in
extract_repository). Only resolved fixes are materialized and
fix-site-indexed.
218 unseen tests pass (incl. 1000 failures after a success not tripping
the guard, and fetch-error regressions), ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The ProcessPool spawned its workers but ordered_extractions blocked the
main thread on the OLDEST future, so while one slow repo finished the
other workers sat idle: live profiling showed 1 of 4 cores busy, ~67%
idle, ~1 attempt/min.
Separate the seed-order acceptance cursor (`pending`: row+decision+key,
no futures) from the worker set (`in_flight`: future->key). Keep up to
`workers` jobs running; journal each result as it COMPLETES via
wait(FIRST_COMPLETED) in any order and refill freed slots, never waiting
on the oldest. A candidate is accepted/rejected only when its result is
in the journal, advanced strictly in seed order -- so the pool runs
out of order while the accepted pool stays byte-identical to a 1-worker
run. Fork-network dedup and the systemic guard still apply at the cursor.
New real-process test proves later extractions are journaled before an
earlier slow one completes (completion order != seed order) while
accepted entries stay in seed order; the old loop fails it. 210 unseen
tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Group 3 was unimplemented, so there was no way to turn the draft into a
frozen pool. Add:
- bisect_guard.py: stages the cache draft and bisects it against the
reserved-repository guard, run as a black box
(pytest -k referenced, stdout/stderr discarded, exit code only; an exit
other than 0/1 aborts as an instrument failure). Drops guard-tripping
candidates and reports the dropped COUNT only -- never a name or URL.
Guard runner is injectable for offline tests.
- freeze.py: writes canonical pool-p1.toml (+ private counterpart,
reusing draft's public/private split and private_digest) and
freeze-p1.json (used-set digest, per-source counts, guard result,
dropped count, label-check counts, freeze digest = sha256 of the
canonical public manifest). A matching retry is a no-op; any changed
frozen artifact requires a new pool id (never an overwrite).
- protocol-p1.md: the frozen scoring/statistics protocol from design
sections 3 and 4.
Offline only; the guard subprocess is the sole subprocess and is mocked
in tests. 42 new tests pass (incl the group-1.3 guard digest check on a
fixture pool); neighbouring unseen suites green; ruff clean. New files
only -- does not touch the journal-pinned draft/extract modules, so a
running draw is unaffected.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Extraction is CPU-bound Python (git-output parsing, HUNK regex,
ast.parse of hundreds of source files per repo). A ThreadPoolExecutor
cannot parallelize it: live profiling showed one thread pegging a core
at 99.7% while three cores sat idle and the others waited on the GIL, so
six "workers" barely beat serial (~0.25 entries/min).
Switch ordered_extractions to a ProcessPoolExecutor with a fork context
(copy-on-write memory; lets the existing in-memory test fakes cross into
children without pickling). Effective workers are clamped to the core
count. The parent still owns the journal, REST (finish_license), decide,
counters, the fork-network dedup, strict seed-order acceptance, the
systemic guard, and the graceful drain. Children run extract_repository
with github=None on picklable args and return their own count deltas
(threads had shared a Counter; processes return counts in the result).
A broken pool is replaced; a collateral worker_crash is a skip outside
the guard, a task-specific extract_worker_failure counts toward it.
Determinism is unchanged and tested with REAL processes: workers=6 ==
workers=1 byte-identical (fork collisions, out-of-order completion,
resume). 208 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Serial extraction at ~147s/repo made a 300-repo pool take days, and
~626 lazy per-file blob faults per clone pushed slow repos past the
240s deadline (a draw reached 21 entries from 347 attempts, 160 of them
extraction_timeout). Both fixed.
Parallel: a bounded ThreadPoolExecutor (--extract-workers, default 6)
runs clones concurrently, each into its own clone with a private counter.
The main thread alone owns the journal, REST (license via finish_license),
counters, and acceptance. Entries are accepted in STRICT SEED ORDER with
the fork-network dedup applied at acceptance, so the pool is byte-
identical to the serial draw (tested: workers=6 == workers=1, including
fork collisions, out-of-order completion, and resume). The systemic guard
and resilient skips are unchanged; a graceful stop drains started work to
the journal for resume.
Batched blobs: repository_changes fetches the fixes together and
Git.materialize fetches all missing blobs for the read revisions in one
`fetch --no-filter --stdin` (rev-list --missing=print), instead of per-
file promisor faults. Content is byte-identical (tested; >=12 fetches
collapsed to 1).
202 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The resilient draw only covered extraction. A per-candidate transport
failure in the decide/eligibility phase -- get_commit on an unresolvable
fix ref raising ValueError("http_failure") (observed: ljharb/shell-quote
commit 1518179) -- still aborted the whole draft, because decide only
caught source.Rejected and eligibility.RepositoryNotFound.
decide now catches that http_failure and returns a "decide_http_failure"
skip, and the run loop feeds decide + extract per-candidate instrument
failures into one systemic guard (25 with zero successes, or >200), so a
single bad commit lookup skips while a broken instrument (e.g. a network
outage failing every lookup) still aborts loudly.
Full unseen draft suite passes (48), ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Individual repositories in the advisory population raise per-candidate
extract.InstrumentFailure during extraction (git_exit_128 on an
unfetchable fix commit, clone_size_cap, and others), and each one
aborted the whole 300-repo draft. The draw is healthy; one bad repo
should not reset it.
extract_repository now also catches extract.InstrumentFailure and
OSError and records them as "extract_<reason>" rejections, so the draw
skips that repo and continues (the reason, e.g. extract_git_exit_128,
shows in the rejections counter). ExtractionTimeout keeps its existing
extraction_timeout skip; any other exception still propagates as a real
abort, so genuine code bugs are not masked.
A systemic guard preserves the instrument rule: run() raises
InstrumentFailure("systemic_extraction_failure") if 25 extraction
attempts fail with zero successes (e.g. git missing, auth broken) or if
extraction failures exceed 200 overall. A handful of bad repos with many
successes never trips it.
190 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
A pathologically slow repository (observed: tensorzero, 183s to clone,
then `git blame -C` past the 180s command limit) raised
subprocess.TimeoutExpired, which Git.run mislabeled as a fatal
InstrumentFailure("git_launch_failed") and aborted the whole 300-repo
draft. The same slowness stalled throughput (a 54-minute no-progress
gap).
Give each clone a per-repo wall-clock budget (--extract-deadline,
default 240s, plumbed Clones -> Git, counting the initial clone). A
timeout -- a command exceeding the remaining budget, or the budget
elapsing -- raises ExtractionTimeout(Rejected), which the existing
rejection handler turns into an "extraction_timeout" skip so the draw
moves to the next candidate. Genuine launch failures and non-zero git
exits stay fatal; timeouts and launch failures are now distinct. The
deadline is recorded on the report only, not the journal inputs, so
changing it does not refuse resume.
181 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
disk_bytes() summed the clone tree with rglob + stat() as a size cap
before every git command. On a blob:none clone, `git show` triggers an
on-demand blob fetch that writes and deletes transient
.git/objects/pack/*.rev files, so a path rglob had just listed could be
gone by the stat() call -> FileNotFoundError aborted the whole draft as
an instrument_failure. The race can hit any candidate, so it was a
non-deterministic draft-killer (observed extracting a real npm repo).
Skip a file that raises OSError mid-walk; the size cap is unaffected.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
The GitHub /advisories endpoint ignores the `page` query parameter: it
paginates with the Link header's rel="next" cursor. The page-number
pager re-fetched the same first 100 rows forever and never terminated,
so enumeration never completed and no draft converged (a run fetched
5137 "pip pages" holding only 100 unique advisories).
Surface the successful response headers from JSONTransport, and walk the
rel="next" cursor per ecosystem until it is absent. The ecosystem /
published>=2019 / bare-number cwes filters are preserved on the first
request. Journal keys stay ecosystem:n (now cursor-walk positions) with
the saved next URL for resume; journal_version 4 refuses older journals.
Live smoke via the production pager: 300 unique advisories pulled in 3
API calls, advancing past the first page. Offline: 170 unseen tests
pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
enumerate_advisories paged the entire global GitHub reviewed-advisory
database (every ecosystem, every year, every CWE) before extraction, so
a 300-repo draft never converged -- a live run reached 21,962 advisories
in 4.5h with extraction never started.
Push the existing candidate() rejections into the query: page per
ecosystem, published>=2019-01-01, and cwes restricted to the seven
families. The cwes filter uses bare CWE numbers (89), which is what the
GitHub advisories API matches -- "CWE-89" silently returns zero. The
whole cwes filter is omitted if the taxonomy ever holds a non-exact
token, so a selectable advisory is never dropped. Journal version 3
(ecosystem:page keys) refuses incompatible older journals.
Live page-1 smoke (pip, npm): 100/100 rows carry a family CWE, none
published before 2019. Offline: 166 unseen tests pass, ruff clean.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
A live G0 sourcing run failed on both: a transient connection/decode error aborted
the whole enumeration (the bare except raised http_failure on the first blip), and
the advisory journal stored full raw advisory JSON and reached 1.77 GB on a near-full
host. Fix: the JSON transport now retries URLError/timeout/ConnectionError/IncompleteRead/
JSONDecodeError through the same attempt-4 backoff as HTTP 5xx/429 (fatal only after the
cap or on a genuinely non-retryable error); enumerate_advisories journals only the fields
resolution/extraction consume via source.reduce_advisory (ghsa id, published/withdrawn,
type, CWE ids, package ecosystem/name for OSV, fix-commit links), with journal_version
bumped to 2 so an old full-format journal is rejected and resume-skip stays intact.
Offline: test_unseen_source/draft/extract 133 passed; ruff + format clean; mypy unchanged
(pre-existing verify.py lambda only).
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Feed the milestone back into the spec and steering. HANDOFF: the open "first demonstrated"
question is RESOLVED -- xiaomusic (path, prefix-check-bypass) demonstrated on the vulnerable
side (3/3 affected, 0/3 safe) and the blind fixed-side search produced 0 false proofs; pair
total $1.22. Records the six-fix chain that enabled it (parity, verify-time-only, fair-share
budget, served-root guidance, oracle prefix-bypass coverage, control-signal feedback) plus the
keystone. tasks.md: implementation note; pilot (6.2/6.3) in progress. spec.json: pilot status.
safety-net steering: proof-based BLOCK mechanism reached first execution-backed demonstration
(feasibility-pilot scope, not release readiness; pilot exit + 73-negative qualification open;
classifier stays triage). search README: status note. Payload-free; demos/boards stay gitignored.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
search-with-proof produced its first demonstrated finding on a real CVE fix-pair end to
end on kube-ax: vuln observes, fixed does not, 0 false proofs, $0.37, 55 calls. Enabled
by the full session fix chain (parity, verify-time-only, fair-share budget, served-root
guidance, oracle prefix-bypass coverage, control-signal feedback). Payload-free record.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Non-blocking advisory hint when the control shows the app didn't serve ${served_root},
so the agent self-corrects its served-directory wiring. Host-side; verdicts unchanged.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
When a path demo does not serve ${served_root}, the keystone control fails
(SideRecord.control_observed False) but the search only reported a generic
inconclusive, so the agent kept making the same wiring mistake. The pilot now
surfaces the affected side's control_observed in the verify result, and the
coordinator appends one advisory, deduplicated "served-root" hint (visible in the
board snapshot the agent reads) telling it to set the served-directory config to the
${served_root} placeholder. Advisory only: verdicts, _end, the oracle, the differential
and the demo validator are unchanged; the hint fires only when control_observed is False.
Tests: coordinator + pilot_run suites 78 passed; ruff clean; mypy unchanged (pre-existing
benchmarks errors only). Host-side change, no image rebuild.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Oracle broadening works in the agent: demo now targets the prefix-sibling vector. Still
inconclusive (0 false proofs, $0.29): the agent set the served-dir env var to its own
value, not ${served_root}, so the sibling canary was outside the served tree. Served-root
wiring is nondeterministic (a prior run wired it). Last-mile agent-reliability gap.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
