Skip to content

Fix nine defects an independent bug hunt found and an adversarial checker reproduced on main dc694994: the credential check's token boundary, case-insensitive scope exclusions, actor identity by princ - #165

Merged
sykosyber merged 2 commits into
mainfrom
syberlabs/5783eda8
Nov 2, 2026

Conversation

@sykosyber

Copy link
Copy Markdown
Member

Objective: Fix nine defects an independent bug hunt found and an adversarial checker reproduced on main dc69499: the credential check's token boundary, case-insensitive scope exclusions, actor identity by principal id, a reconciled pull request's destination, provider stdin under the deadline with a tree kill, tampered hidden-check counts, unknown-metric specs and huge thresholds. Each fix carries its own regression test and one GAPS entry records what was closed and what remains.

Class: R
Agent-platform: claude@fable-5.1
Admitted-by: sykosyber
Reviewed-by: rules@2 (critic check critic.rules: high 0, medium 0, low 0)
Evidence: tested

Accepted in SyberLabs thread 5783eda8-3306-4e59-ae59-5fbe5e7de120 as candidate c3 (commit 8443275a8eef0c60f7e7eb8eca9ae2965e53104b).
Host checks on that tree: conformance passed, critic.rules passed, scenarios passed, tests passed.

…tring hunk, inside the contract's 30-file scope)

SyberLabs-Thread: 5783eda8-3306-4e59-ae59-5fbe5e7de120
SyberLabs-Candidate: c3
SyberLabs-Provider: claude@fable-5.1
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-02T09:30:56.742082Z 8443275 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8443275a8e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

"request_digest": {"type": "string", "pattern": "^[0-9a-f]{64}$"},
"response_digest": {"type": "string", "pattern": "^[0-9a-f]{64}$"},
"idempotency_key": {"type": "string", "minLength": 1},
"destination": {"type": "string", "minLength": 1},

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Version this reconciliation proof change

When recovering a GitHub pull request after a lost response, the emitted proof now includes destination. A client still validating sdk.syberlabs.space/v0alpha1 against the prior schema will reject it because the proof disallows additional properties. Keep the frozen v0alpha1 wire shape or make this a versioned protocol change.

AGENTS.md reference: AGENTS.md:L24-L25

Useful? React with 👍 / 👎.

@sykosyber

Copy link
Copy Markdown
Member Author

Scope: owner direction in session 2026-11-02 (SyberFactory, packet hunt-fixes-1: the nine defects an independent bug hunt found and an adversarial checker reproduced on main dc69499491).

Packet hunt-fixes-1, thread 5783eda8, candidate c3 (head 8443275a8eef0c60f7e7eb8eca9ae2965e53104b, tree 9fb928905418315cf7afc67814628b7a394c1788), base main at dc69499491. Built through contract v5 (thread_setup_v5.py --base main --max-dollars 5), class R, unattended. Three candidates: c1 failed the thread's tests check (two failures, both from the server's actor rename, below); c2 fixed them but was OUT OF SCOPE: max_files (31 files against the contract's 30); c3 is c2 without the docstring-only hunk in syberwork/core.py. Three syberlabs run --once passes: the first checked c1 (refused, candidate_check_failed:tests), the second checked c3 acceptable, the third accepted it (refs/heads/syberlabs/5783eda8 at 8443275a8e). No model and no paid call was made; no GitHits call was made. Every test uses stand-ins and loopback.

What changed, defect by defect

  1. INT-1 (medium), credential token boundary. syberlabs/query_guard.py (secret) and syberlabs/eval/gate.py (credential_observed) missed a token after a word ending in a credential prefix: task-ghp_…, disk-sk-ant-api03-…, see ask-AKIA… and key_sk-ant-api03-… all passed. Cause: the regex found sk- inside task-, the word-start rule rejected that match, and finditer resumed after the whole token, so the real prefix behind the hyphen was never seen; key_sk-ant- failed on the _ boundary. Fix: one shared syberlabs/_patterns.py: credential_prefixes (the hook process loads _patterns.py beside the guard, so it gets it too). A prefix counts wherever it is not glued to a preceding letter or digit, and a rejected start resumes one character on. Words ending in sk still pass: risk-assessment, task-list-manager, disk-cleanup-tool, ask-anything-bot, flask-sqlalchemy-utils, mask-generation-model, and the gate's existing task-pricing-run12345678 ask-ant-x. Tokenization only: no base64/hex detection (pilot freeze SW-08). Tests: test_query_guard.Rules.test_a_credential_after_a_word_ending_in_a_prefix_is_still_a_credential (six hidden tokens, seven pass cases, and the helper never returns the key); test_eval_score.GateRules.test_credential_observed extended with the glued token.
  2. INT-2 (medium), case-insensitive scope. syberlabs/evolution.py: in_scope/_matches compared prefixes case-sensitively. Fix: CASE_INSENSITIVE_PATHS, probed once on the package's own path (the path spelt in its letters' other case is the same file; a path without letters falls back to os.path.normcase); where true, entries and paths casefold on both sides, inclusion and exclusion alike; elsewhere they compare as written. The kernel has no second scope check: syberwork.core.record_candidate recomputes scope_violations through evolution.candidate_record, which calls this in_scope, so one rule governs both. Documented in docs/EVOLUTION.md and the docstring. Tests: test_evolution.ScopeCase (three tests: the folded rule, the as-written rule, and the host probe).
  3. INT-3 (medium), actor identity. syberwork/server.py passed user["name"] as the actor to every kernel call, and syberwork/builder_http.py: _actor did the same, so two principals named bot were one actor to approve, commit, evaluation and signoff. Fix: builder_http.actor_of(user) returns the principal id for a registered principal and the configured name only for a users.json entry (whose keys are distinct); the server's do_POST uses it for all 19 kernel calls (the local variable is actor now); /api/me keeps the display name. The one exception is the witness request: request_witness(…, actor, principal) records both in a trace and compares neither, so it keeps the display name as its label beside the principal id (test_console pins ('lead', 'person-1')). The G6 static lint's mutation anchor in tests/test_g6_bridge_mapping.py follows the rename (name, roles → actor, roles); it is a changed anchor, not a new test. No kernel comparison changed: the kernel compares the string it is given, which is now the id. Test: test_server.Principals.test_two_principals_with_one_display_name_are_two_actors (bot-2 cannot commit bot-1's proposal: 409 actor_mismatch; bot-1 can; every recorded actor is bot-1; /api/me shows bot with principal_id bot-2).
  4. F1 (medium), a recovered pull request's destination. syberlabs/session.py: reconcile built the verified proof without the destination, so after a lost response metrics.work_items read the thread as never opened and delivered=False while the pull request was open; build.recover's receipt named the action and a retention receipt had destination: None. Fix: the proof carries destination when the executor's status proof has one (spec/reconciliation.schema.json: optional proof.destination); retention reads it; recover's receipt uses it as ref; metrics already read the proof. Test: test_publish.test_pull_request_is_created_once_even_when_the_response_is_lost extended: receipt ref == github:acme/shop, the proof's destination, and work_items says opened #1, delivered unknown ("pull request open"). Conformance's lost_ack trace is unchanged (its executor's proof has no destination, so no key is added).
  5. F2 (medium), stdin under the deadline. syberlabs/providers.py: run_json wrote the whole payload before its deadline loop, so a child that never read stdin held the host in write until it exited and the empty stdout then read provider_shape. Fix: providers.feed_stdin writes on a daemon thread and the loop starts at once; the error at the deadline is provider_timeout. syberlabs/claude_code.py: run_stream had the same shape; the change there is the import and one line replacing the five-line write block (two hunks, far from the guard lane). Tests: test_providers.RunJson.test_a_child_that_never_reads_its_input_is_stopped_at_the_deadline (200 KB payload, child sleeps 8 s, timeout 1: provider_timeout in under 5 s); test_claude_code.Timeout.test_a_prompt_the_stand_in_never_reads_does_not_hold_the_host_past_its_deadline.
  6. F3 (medium), tree kill. run_json killed the direct child only. Fix: checks._stop (killpg on POSIX, taskkill /T /F on Windows), imported inside the timeout branch because checks imports critic, which imports providers. Test: test_providers.RunJson.test_a_timeout_ends_the_grandchild_too (the child writes its grandchild's pid, both sleep; the grandchild is gone within 10 s of provider_timeout), using test_checks.alive.
  7. F4 (low), tampered hidden parts. syberlabs/eval/score.py: run_metrics divided the passed count by the total whether or not quality["trusted"]. Fix: hidden_parts_passed is None when the hidden copy was tampered. Test: test_eval_score.Scoring.test_a_tampered_hidden_copy_leaves_hidden_parts_passed_unknown, which also pins the metric keys to spec.METRIC_NAMES.
  8. F5 (low), unknown metric names. syberlabs/eval/spec.py: check accepted any primary_metric.name, secondary_metrics[].name and decision_rule.metric. Fix: spec.RUN_METRICS (the seven of score.run_metrics) + syberlabs.metrics.METRICS = spec.METRIC_NAMES; an unknown name is refused with the known names listed. Test: test_eval_adversarial.PreRegistration.test_a_metric_no_scorer_computes_or_a_threshold_no_float_holds_is_not_a_valid_spec. Every registered spec under evals/ names only known metrics.
  9. F6 (low), a huge threshold. 10**400 passed {"type": "number"} as an integer and report.decide raised OverflowError. Fix: spec.check requires a number a finite float holds; report._threshold fails closed with Rejected("threshold_invalid") for 10**400, infinity, NaN, a bool, a string or None, and 10**300 still decides. Tests: the spec test above and test_eval_score.Statistics.test_a_threshold_no_finite_float_holds_is_refused_not_raised.

spec/GAPS.md entry 65 records what closed and what stays open; it was the next free number on main at propose time (another in-flight packet also appends 65, the expected trivial conflict for the merge script). Docs touched in one sentence each: docs/EVOLUTION.md, docs/CLAUDE_CODE.md, docs/EVAL.md (three rows), docs/INDUSTRIAL_CELL.md, docs/METRICS.md.

Files (30): code syberlabs/_patterns.py, syberlabs/query_guard.py, syberlabs/eval/gate.py, syberlabs/evolution.py, syberwork/builder_http.py, syberwork/server.py, syberlabs/session.py, syberlabs/retention.py, syberlabs/build.py, syberlabs/providers.py, syberlabs/claude_code.py, syberlabs/eval/score.py, syberlabs/eval/spec.py, syberlabs/eval/report.py, spec/reconciliation.schema.json; tests tests/test_g6_bridge_mapping.py (anchor), tests/test_query_guard.py, tests/test_eval_score.py, tests/test_eval_adversarial.py, tests/test_evolution.py, tests/test_server.py, tests/test_publish.py, tests/test_providers.py, tests/test_claude_code.py; docs and spec/GAPS.md as above. tests/test_eval_core.py, eval/runner.py untouched (#161); claude_code.py two small hunks (guard-lane-2).

Commands and results

  • Targeted, local (Windows 11, CPython 3.12.6): first pass test_query_guard, test_evolution, test_eval_score, test_eval_adversarial.PreRegistration, test_server, test_gaps, test_providers.RunJson, test_publish, test_claude_code.Timeout, test_session, test_forget, test_reconciliation, test_builder_http, test_cell: all OK after two test-side fixes and one real one (decide must also refuse a string threshold; see F6). Wide pass over test_checks, test_metrics, test_eval_runner, test_eval_core, test_build_thread, test_providers, test_claude_code, test_githits_adversarial, test_work, test_admission, test_attest_reconcile_map, test_evolve, test_jev, test_governor, test_sdk_surface, test_package, test_eval_adversarial, test_journal, test_critic: Ran 344 tests in 603.9s, 343 OK (skipped=3, expected failures=4); the one error is a module name I mistyped on the command line (tests.test_retention does not exist; retention is test_forget, OK above).

  • Local: python -m conformance.run: matched 23 golden traces. python benchmarks/run_scenarios.py: scenarios: 11 replayed, exit 0, digest b41f7cbb… (unchanged).

  • Full suite, GitHub Actions, scratch branch syberlabs/5783eda8-ci (e173521a, tree 9fb92890, the candidate's tree): https://github.com/SyberLabs/SyberWork/actions/runs/41417925336, success after one rerun of the failed jobs. First attempt: test (3.11), test (3.12), test (3.13), compose and G8 strict passed; windows failed on one test I did not touch, test_githits_pace.CrossProcess.test_two_concurrent_launcher_examples_are_spaced_recorded_and_other_ops_unpaced (1.7037 not >= 1.75: two launcher arrivals 46 ms closer than the 2 s interval minus the 0.25 s tolerance; a timing assertion on the runner, on a path that imports nothing this packet changed), and G8 Existing suite only aggregates the legs. The rerun of the failed jobs passed (windows, G8); all other legs kept their first result. Every leg: Ran 1273 tests, OK (Linux skipped=2, Windows skipped=26, expected failures=4). The PR's own checks on Fix nine defects an independent bug hunt found and an adversarial checker reproduced on main dc694994: the credential check's token boundary, case-insensitive scope exclusions, actor identity by princ #165 at 8443275a8e all pass on the first attempt, windows included (gh pr checks 165: G2, G3, G4, G6, G8, G8 strict, compose, policy, test 3.11/3.12/3.13, windows).

  • Bite, GitHub Actions, scratch branch syberlabs/5783eda8-bite (5a023f10: only the nine test files on dc69499491): https://github.com/SyberLabs/SyberWork/actions/runs/41417861525, failure, as intended. test (3.12): Ran 1273 tests, FAILED (failures=13, errors=14, skipped=2, expected failures=5); every failure and error is a new or changed test and nothing else fails: the guard test (errors: no _patterns.credential_prefixes; failures on the four hidden tokens), test_credential_observed, the three ScopeCase tests (no evolution.CASE_INSENSITIVE_PATHS), Principals, the lost-response publish test, both RunJson tests, the run_stream deadline test, the tampered-parts test, the threshold test, the spec test, and the G6 anchor. windows, 3.11 and 3.13 fail the same way; compose passes.

  • Thread checks on c1 (first run --once): critic.rules, scenarios and conformance passed; tests failed: Ran 1273 tests, FAILED (failures=2): test_console.ConsoleWitnessRequest.test_each_request_is_counted_and_traced_with_its_actor (the witness trace's actor read person-1 instead of lead) and test_g6_bridge_mapping.G6StaticLint.test_lint_fails_closed_on_seeded_violations (its anchor …, name, roles) no longer in server.py). Both fixed as under item 3.

  • Thread checks on c3 (30 files, 76,398 diff bytes):

    Check State Time Output
    conformance passed 12.9 s matched 23 golden traces
    critic.rules passed 1.9 s no findings
    scenarios passed 0.5 s scenarios: 11 replayed, exit 0, digest b41f7cbb…
    tests passed 480.1 s Ran 1273 tests, OK (skipped=26, expected failures=4) (host-run shards, Windows)
  • certify c3: result 8443275a8e (tree 9fb9289054) on repo-change.v5, class R; four checks passed; chain head 3aafa7d5f5 at 21 events; written .pub equal to the factory's key (64 hex characters; the files differ only in line ending). Key 3312866a18 (the rotated key): the derived public prefix of the seed read from the user environment equals factory-key-current.pub before signing.

  • verify against the factory's .pub with --repo .: signature ok ed25519 3312866a18; repository ok (base dc69499491 and result present, trees match, 30 changed paths and diff digest match, target ref at 8443275a8e); journal ok (21 of 21 events certified, head 3aafa7d5f5); verdict pass.

Claim ceiling: implemented and tested (locally and on CI: Linux 3.11–3.13, Windows), against stand-ins. Not measured in a live run, not deployed.

Known limitations

  • The register_principal docstring sentence was dropped to stay inside the contract's 30-file scope; the rule is stated in builder_http.actor_of and docs/INDUSTRIAL_CELL.md.
  • The case rule is probed on the filesystem holding syberlabs/evolution.py, not the repository's: a repository on a volume of the other kind compares the wrong way (GAPS 65 (a)).
  • A users.json name equal to a principal id is one actor; both reach the kernel as plain strings (65 (b)).
  • The kernel's own syberwork.core.reconcile (destination records) still writes no destination into its proof (65 (c)); the brief named session.py.
  • feed_stdin's thread ends on its own once the child is gone; a write blocked on a full pipe cannot be interrupted earlier (65 (d)). _stop has the checks' limits: a grandchild that detaches into its own session can survive (65 (e)).
  • spec.RUN_METRICS is pinned beside score.run_metrics and held equal by a test, not derived (65 (f)).
  • Secondary metric names are validated too, beyond the brief's primary and decision metric; every registered spec passes.

Unresolved questions

  1. Does the owner want the kernel's reconcile proof to carry a destination too (65 (c))? It is a destination-record path, outside this packet's files.
  2. The scratch branches syberlabs/5783eda8-ci and -bite are left for the orchestrator to delete.

Next permissible packet: unchanged; the pilot freeze items (SW-01, 03, 04, 05, 07, 08, 10) were not touched.

@sykosyber

Copy link
Copy Markdown
Member Author

Independent verification of #165 (hunt-fixes-1, c3 8443275a8e): pass with findings

Read-only claim check on a fresh clone (tree 9fb9289054, 30 files vs dc69499491), Windows 11, CPython 3.12. No model, GitHits or paid call. Nothing merged, pushed or edited.

Re-run of the nine repros (each now fails closed)

# Repro Result
INT-1 task-ghp_…, disk-sk-ant-api03-…, see ask-AKIA…, key_sk-ant-api03-… through query_guard._credential and gate._secret_matches all blocked / flagged; helper returns prefixes only, never key material
INT-1 FP risk-assessment, task-list-manager, disk-cleanup-tool, ask-anything-bot, flask-sqlalchemy-utils, mask-generation-model, task-pricing-run12345678 ask-ant-x + three more (helpdesk-ticket-router-20251102, ramdisk-loader-v2-build-12345678, key_rotation_policy_v3_2025) all pass, both sides
INT-1 hook real launch python -I <abs>/syberlabs/query_guard.py <dir> via hook_command + hook_environment the four tokens exit 2 (rule: secret), benign queries exit 0: _beside("_patterns") resolves under -I
INT-2 in_scope({"paths": ["."], "exclude": ["tests"]}, "Tests/conftest.py"), .SyberLabs/x, .GIT/config on this host (CASE_INSENSITIVE_PATHS=True) all False; with the flag forced False they are admitted, as documented for POSIX
INT-3 test_server.Principals bot-2 commit of bot-1's proposal: 409 actor_mismatch; every actor bot-1; /api/me name bot, principal_id bot-2
F1 test_publish…response_is_lost receipt ref == github:acme/shop, proof destination, work_items opened #1, delivered unknown ("pull request open")
F2 run_json, 200 KB payload, child sleeps 8 s, timeout 1 provider_timeout after 1.25 s (was provider_shape after 8 s)
F3 child spawns a sleeping grandchild, timeout 2 provider_timeout; grandchild gone at 2.4 s (test_checks.alive)
F4 run_metrics with trusted=False, 2/2 passed hidden_parts_passed=None; trusted=True → 1.0; keys == spec.METRIC_NAMES (16 = 7 + 9)
F5 spec.check with bogus_metric as primary + decision metric refused with the known-names list; evals/e5-single-vs-repair/spec.json still validates (0 errors), spec_digest == registration.spec_digest (7625340097b9…), spec.registered OK
F6 report.decide with 10**400, -(10**400), inf, NaN, True, "0.5", None each Rejected('threshold_invalid'); 10**300 decides (holds=False); spec.check refuses 10**400 without OverflowError

Tests and checks

  • test_query_guard test_eval_score test_eval_adversarial test_evolution test_server test_publish test_g6_bridge_mapping test_console test_gaps test_eval_core: Ran 224, OK (skipped=2, expected failures=4). test_providers test_claude_code: Ran 64, OK.
  • python -m conformance.run: matched 23 golden traces (lost_ack unchanged). benchmarks/run_scenarios.py: 11 replayed, exit 0, digest b41f7cbb… (unchanged).
  • gh pr checks 165: all 12 pass (G2, G3, G4, G6, G8, G8 strict, compose, policy, test 3.11/3.12/3.13, windows).
  • Certificate hunt-fixes-1-5783eda8.cert.json vs factory-key-current.pub, --repo = fresh clone: signature ok ed25519 3312866a18; repository ok (base and result present, trees match, 30 paths and diff digest match, refs/heads/syberlabs/5783eda8 at 8443275a8e); journal skipped (no thread journal under a fresh clone's .syberlabs; the canon checkout lacks the result commit). The handoff's "journal ok 21/21" is from the factory's own working copy and is not independently re-checked here.

Pilot freeze

Confirmed untouched: no retry/back-off, base64/hex, reopened/pr_reopened, lookups_blocked, shard, prompt-builder or metric-source change anywhere in the diff; syberlabs/metrics.py, githits*, settlement not in the file list. query_guard.py hunks are tokenization only (_patterns.credential_prefixes: prefix not preceded by alnum, resume at start+1), no encoding detection. session.py:455-462 adds destination only inside the state == "applied" branch; the reopened case is not in this path.

Main interplay (trial merge onto main 8a5e21f)

Findings

Low

  1. syberlabs/_patterns.py:22-31 credential_prefixes is quadratic on a whitespace-free run of key-class characters with glued prefixes. Each rejected start resumes one character on and the greedy [A-Za-z0-9_\-]{8,} re-consumes the rest of the run. Measured on "task-" * n: 48 KB 0.34 s, 97 KB 1.28 s, 195 KB 5.70 s (×4 per doubling; the old finditer pass was linear, <1 ms). The query guard is unaffected (max_chars 400). The eval gate is: gate.py:445-448 credential_observed reads every evidence file whole, uncapped (_evidence_files: every non-derived, non-binary file in the run folder), so a ~1 MB such run in evidence costs minutes and 10 MB hours of scoring. Fail-slow, not fail-open; nothing is missed. Suggest a (?<![A-Za-z0-9]) lookbehind so one finditer pass gives the same rule linearly, or a cap on the scanned bytes. Space-separated text (195 KB of task-list ) scans in 2 ms.
  2. syberlabs/evolution.py:142-143 casefold() is wider than any filesystem's case rule. _matches("STRASSE/x", "straße") is True; NTFS and APFS do not fold ß→ss. Over-match only: exclusions fail closed, paths can admit a distinct file. Edge case, non-ASCII scope entries only.

Informational
3. INT-3 count: 18 call sites in do_POST now pass actor (17 work.* + _preview); request_witness (server.py:372) is the 19th and keeps user["name"] as a trace label beside principal_id, compared nowhere (test_console pins ("lead", "person-1")). The handoff's "all 19" counts the one it names as the exception. No remaining user["name"] reaches a kernel comparison; /api/me (server.py:198) keeps the display name. GAPS 65(b): register_principal has no HTTP route and no product caller outside syberwork/core.py; the id/users.json-key collision needs the cell's administrator to pick a colliding id — configuration, not a client-reachable path.
4. INT-2 probe: _case_insensitive_filesystem(__file__) (evolution.py:76-85) is sound on POSIX (normcase is identity there, so a case-insensitive volume is detected only by the samefile probe, which is right). The known limit (65(a)) cuts both ways: a case-insensitive host with a case-sensitive repository directory (WSL, fsutil per-directory flag) over-includes (Tests/x admitted by paths: ["tests"]), not only under-excludes.
5. F2 feed_stdin (providers.py:213-226): BrokenPipeError ⊂ OSError, so the tuple is redundant but correct; on Windows the write to a dead child's pipe raised OSError and was swallowed (the 1.25 s repro). _stop ends with proc.wait(), so no zombie. A grandchild that inherits the stdin read end keeps the daemon writer, its pipe handle and the payload buffer alive past writer.join(5) until it exits (65(d)/(e) territory).
6. F3 lazy import: the cycle is real (checks.py:59 imports critic; critic.py:45 imports providers.model_usage); a top-level import in providers would break import syberlabs.providers first. claude_code.py importing checks._stop at top is fine because providers does not import checks at top.
7. syberwork/core.py is untouched (register_principal docstring unchanged); the actor rule is stated in builder_http.actor_of (builder_http.py:28-36) and docs/INDUSTRIAL_CELL.md.

Nothing refused during this check.

…(GAPS entry renumbered 65 -> 66)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@sykosyber
sykosyber merged commit 91251b5 into main Nov 2, 2026
25 checks passed
sykosyber pushed a commit that referenced this pull request Nov 4, 2026
#165 added the optional proof.destination

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant