admin/PlantAI-dev
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity
Files
master
PlantAI-dev/.superpowers/sdd/task-3-definitive-verification.md
T
potaser f745ee9d7e Initial commit: PlantAI-dev
2026-08-03 18:42:47 +08:00

9.6 KiB
Executable File
Raw Permalink Blame History

Task 3 definitive verification

Verdict

PASS

  • Critical: 0
  • Important: 0
  • Minor: 0
  • Checkpoint recommendation: APPROVED for Task 3 only

Fresh definitive verification found no remaining Task 3 defect. The current production behavior closes the two Important and one Minor findings in the latest final spec/code-quality and anti-pattern reviews, as well as every earlier Task 3 finding. All automated and direct production probes ran with live sockets and DNS disabled.

Material and patch surface reviewed

Read before verification:

  • AGENTS.md and CLAUDE.md;
  • .superpowers/sdd/task-3-brief.md;
  • updated .superpowers/sdd/task-3-report.md and .superpowers/sdd/task-3-review-package.md;
  • .superpowers/sdd/progress.md;
  • the latest authoritative reports .superpowers/sdd/task-3-final-review.md, .superpowers/sdd/task-3-final-antipattern.md, and .superpowers/sdd/task-3-final-verification.md; and
  • the prior Task 3 verification, anti-pattern, and spec/code-quality reports.

This workspace has no functional Git repository (.git contains only info/), so the review-package baselines were used. Current baseline/current diff and complete-file surfaces were inspected:

/tmp/ppi-id-resolution-task3-baseline/run.py
  -> skills/ppi/run.py                         1,163 diff lines
/tmp/ppi-id-resolution-task3-baseline/pipeline_backend.py
  -> apps/web/pipeline_backend.py                146 diff lines
/tmp/ppi-id-resolution-task3-baseline/test_ppi_string_policy.py
  -> tests/unit/test_ppi_string_policy.py        100 diff lines
gfa_core/string_rate_gate.py                     148 complete lines
gfa_core/ppi_string_policy.py                    297 complete lines
tests/unit/test_ppi_id_resolution.py            1,360 complete lines

The two authorized backend positive image fakes and the authoritative PPI selection test in tests/unit/test_pipeline_backend_execution_env.py were also inspected. The production patch remains limited to the documented Task 3 surface and parent-authorized policy/backend expansions. No Task 4 Reset/CLI ID-map discovery or Task 5 contract/diagnosis/frontend work was pulled into this checkpoint.

Fresh socket-blocked automated verification

Each pytest command ran inside a fresh Python process after replacing these four network entry points with a function that raises AssertionError("live network blocked by definitive Task 3 verifier"):

socket.socket.connect = blocked
socket.socket.connect_ex = blocked
socket.create_connection = blocked
socket.getaddrinfo = blocked

The forked multiprocessing rate-gate tests inherit this block.

Focused Task 3 suite

/home/xia-lab/miniconda3/envs/test4py311/bin/python [socket-blocked pytest]
  tests/unit/test_ppi_id_resolution.py
  tests/unit/test_ppi_string_policy.py -q

collected 123 items
123 passed in 0.98s
exit 0

Relevant backend image/PPI selection

/home/xia-lab/miniconda3/envs/test4py311/bin/python [socket-blocked pytest]
  tests/unit/test_pipeline_backend_execution_env.py
  -k 'string_network_image or ppi' -q

collected 32 items / 5 deselected / 27 selected
27 passed, 5 deselected in 6.05s
exit 0

Combined regression

/home/xia-lab/miniconda3/envs/test4py311/bin/python [socket-blocked pytest]
  tests/unit/test_id_resolution.py
  tests/unit/test_ppi_id_resolution.py
  tests/unit/test_ppi_string_policy.py
  tests/unit/test_pipeline_backend_execution_env.py -q

collected 167 items
167 passed in 7.03s
exit 0

Syntax compilation

/home/xia-lab/miniconda3/envs/test4py311/bin/python -m py_compile \
  gfa_core/ppi_string_policy.py \
  gfa_core/string_rate_gate.py \
  skills/ppi/run.py \
  apps/web/pipeline_backend.py

exit 0; no output

Pytest emitted the pre-existing environment-level RequestsDependencyWarning for installed urllib3/chardet versions. It did not affect collection or assertions.

Fresh direct production probe

/tmp/task3_definitive_probe.py imported the current production modules, blocked the same four socket/DNS entry points, used only temporary files and local fake clients, and exited 0. Its final terminal conclusions were:

PASS species self-resolution and stable URL authority
PASS mapping queryIndex/echo/taxon/stringId shape and 404/empty semantics
PASS submitted-ID taxon, endpoint membership/shape, and score threshold
PASS redirect/final authority, typed status, and empty-vs-failure semantics
PASS direct/projected raw evidence provenance
PASS empty/partial Ortholog, 24->12 provenance, and wrapper dispatch
PASS backend authoritative skill selection; no legacy run_ppi_v2 fallback
PASS strict image status/final URL, redirect blocking, and shared caller/gate
PASS rate-gate pacing and nonfinite/corrupt/unreadable fail-closed safety
TASK3_DEFINITIVE_DIRECT_PROBE_PASS sockets/DNS blocked

Two preliminary probe executions stopped on verifier-harness expectation-type mismatches (StringApiError versus ValueError); only the /tmp probe was corrected. The authoritative full rerun above started from the beginning and exited 0. No production or repository test file was changed.

Historical-finding closure evidence

Historical finding Fresh definitive evidence
Empty/unknown species silently defaulted to Arabidopsis; declared codes were unreachable Every declared STRING_SPECIES_MAP canonical code self-resolved to its exact taxon. Empty and unknown species raised before HTTP.
Stable address authority was fail-open Genuine authoritative STRING HTTPS hosts were accepted. HTTP, suffix-deception, credentials, port, empty-label, double-slash, extra-path, and query-bearing bases were rejected.
Redirects could resend identifiers; final response authority was ignored Scientific 307 and optional-image 307 probes made only configured requests with allow_redirects=False; no Location was followed. Evil final response URLs were rejected before scientific JSON parsing or image writing.
Mapping attribution/shape/taxon/STRING-ID contradictions were accepted Validated queryIndex, echoed query, strict scalar taxon, nonblank preferred name, and selected-taxon STRING-ID prefix all failed closed under adversarial rows. Mapping 404 and 200 [] remained zero resolution.
Submitted network identifier could carry the wrong taxon 3702.FOREIGN under selected taxon 3562 raised a typed protocol error before any HTTP call.
Returned endpoints could fall outside the exact submitted set An A-B response for an A-only submission raised a typed protocol error. A valid A-B response required both exact submitted STRING IDs.
Returned score could be below required_score Score 0.5 under required_score=700 raised a typed protocol error; score 0.8 passed.
Missing/blank/foreign endpoint identity, taxon, or nonfinite score was accepted Missing/blank STRING IDs and names, foreign endpoint prefixes, boolean taxon, and NaN score all raised typed protocol errors.
Malformed status escaped the typed boundary; network failure collapsed to empty science Boolean/missing status failed closed. Network 200 [] remained valid empty science; network 503 remained a typed HTTP failure.
Direct raw evidence falsely claimed an Ortholog relationship Direct raw resolution emitted canonical_member_input with an empty Ortholog ID. Projected raw resolution emitted ortholog_relationship with the exact Ortholog ID.
Empty/comment-only supplied Ortholog silently switched to direct mode Both inputs produced exact ID-resolution metrics 2 input / 0 relationships / 0 resolved / 2 unresolved, with zero mapping and network calls.
Missing canonical members disappeared from denominators A partial Ortholog file produced exact metrics 2 input / 1 relationship / 1 resolved / 1 unresolved.
Relationship provenance was lost during deduplication A production PpiSkill.run() probe retained 24 TSV relationship rows, collapsed to 12 unique reference queries and 12 selected STRING IDs, with exact 5 raw plus 7 GeneID-fallback query metrics.
Signature inspection could select a legacy resolver A subclass wrapping analyze_network(self, *args, **kwargs) still ran deterministic resolution and submitted only the selected STRING ID.
Backend could reactivate apps/cli/run_ppi_v2.py A dynamic missing-skill probe exposed the legacy file as present, but run_stage("ppi", ...) failed closed, launched nothing, and returned the authoritative-skill/legacy-disabled message. Static inspection found no production run_ppi_v2.py reference.
Optional image responses without explicit status/final URL were accepted Missing status, missing URL, evil final URL, and 307 responses were rejected across all three attempts. A positive response required explicit integer 2xx, authoritative final URL, redirects disabled, PlantAI, and the shared gate.
Rate gate could fail open on unsafe state Deterministic pacing slept exactly the remaining interval. Nonfinite/negative intervals, nonfinite clock, empty/NaN/infinite/nonnumeric timestamp, and unreadable timestamp state all failed closed; the focused suite also freshly passed both real multiprocessing serialization tests.

Scope and residual validation

  • Verdict scope is Task 3 only.
  • No live STRING request, backend service/Celery execution, or full real-job run was performed. Those remain Task 6 validation; socket blocking proves this verification stayed offline.
  • No service or Celery restart was needed.
  • No diagnosis record was updated because diagnosis is Task 5 scope.
  • No production or repository test source was edited. This report is the only repository file created by this definitive verification.
Reference in New Issue View Git Blame Copy Permalink