Skip to content

πŸ›πŸ§‘β€πŸ’» - Progress scope assessment rejects an API requirement with scope clause limit in generated issue plan metadataΒ #828

Closed

Description

@efraespada

Is there an existing issue for this?

  • I have searched the existing issues.

The exact diagnostic search returned only unrelated closed setup issue #274. Related closed #774 improved rejected-progress diagnostics; this report concerns the new, explicit scope-clause-limit failure and recovery, not an absence of those diagnostics.

Which actions are affected?

Commits

Which platforms are affected?

Linux

Description

A completed progress scope result is rejected with scope-clause-limit at $.lines for source L119. Copilot retains existing progress and fails the consumer's commit verification before publishing a valid assessment, while the independent project CI Check passes.

The reported source anchor is inside the Action-maintained hidden recommendationState.implementationPlan.steps details in the consumer issue body. That is an observation about the anchor, not proof that metadata inclusion, the scope bound or the model caused the failure. Please investigate which input and returned clause count exceeded which bound, whether derived recommendation metadata is intended to become progress scope, and the supported recovery for this coherent issue. Keep fail-closed validation and prior progress.

This is one observed completed-result rejection. No deterministic reproduction rate, provider outage, malformed source requirement or causal link to application behavior is asserted. The diagnostic is now much more useful than the earlier #774 case; the remaining problem is reaching a valid bounded scope without changing consumer requirements or weakening acceptance.

Reproducing the issue

  1. Consumer issue: βœ¨πŸ§‘β€πŸ’» - Implement Home Assistant identity and SQLite platform persistenceΒ domus#6; draft PR: Implement HA identity, SQLite persistence and verified restorationΒ domus#7.
  2. Copilot - Commit run: https://github.com/vypdev/domus/actions/runs/41168137540, job https://github.com/vypdev/domus/actions/runs/41168137540/job/124440053296.
  3. Exact comparison: before 16999cd3635a8f475c04187e7e1e3c7caec83e86; after 780281e842142ec16c3c214cc7236582e59aa0e5. Run attempt 1, push event.
  4. The admitted codex-cli 0.166.0 agent completes, then scope validation rejects $.lines with reason scope-clause-limit and source L119. No verified requirement ID is available (requirement none).
  5. Native Copilot / Verification reports lifecycle blocked, error agent.failed, retryable Yes, reference fb65aecf-0fd2-457c-b788-551d7ce78afe. Existing progress is retained.
  6. At inspection, raw issue L117 is the generated step title "Expose the versioned identity API"; L119 is its details string: "Add validated FastAPI identity, household, member, and policy endpoints with safe error envelopes and no credential persistence or diagnostic leakage."
  7. Independent exact-head CI Check is green: https://github.com/vypdev/domus/actions/runs/41168141080. Copilot - Branch Sync and Copilot - Pull Request also succeeded. Draft review is neutral/skipped and is not treated as approval.
  8. No Copilot source, installed workflow/configuration, issue scope, coverage gate or managed branch was changed to bypass this failure.

copilot Version

vypdev/copilot@v3 resolves in the failing job setup to 3e7ceea. Agent runtime: codex-cli 0.166.0. Exact source revision is authoritative; no unverified package version is inferred.

Relevant Log Output

Download action repository 'vypdev/copilot@v3' (SHA:3e7ceea08e6131360e026937d1c0cc1b120472e0)
BEFORE_SHA: 16999cd3635a8f475c04187e7e1e3c7caec83e86
AFTER_SHA: 780281e842142ec16c3c214cc7236582e59aa0e5
Agent execution admitted (codex codex-cli 0.166.0).
Agent execution completed.
Progress scope result was rejected (scope-clause-limit; field $.lines; requirement none; source L119). Existing progress was retained.
Error code: agent.failed
Action: Inspect the original requirement and response schema.
Details: phase scope, reason scope-clause-limit, field $.lines, requirement none, original issue line L119.
Retryable: Yes
Reference: fb65aecf-0fd2-457c-b788-551d7ce78afe

Additional context and comments

The consumer increment has 490 local tests (375 backend, 65 tooling, 39 architecture fixtures, 11 Node policy), 100% executable backend statement/branch coverage, unchanged 29 installed Copilot resource hashes, original SDD/catalog checks and matching green application CI. This does not establish scope-agent correctness or full M1 acceptance; the feature intentionally remains a draft.

Suggested investigation/acceptance:

  • Pin the original issue snapshot and distinguish human normative scope from generated or excluded metadata before choosing a fix.
  • Exercise a legitimate API requirement with several nouns/invariants, long and short human descriptions, hidden generated recommendation JSON, repeated source anchors, and adversarial over-splitting.
  • Keep clause/count/input/output bounds explicit and tested; reject malformed or unsupported results without partial progress publication.
  • If a bounded repair/retry is supported, preserve the complete scope, source identity, old progress and exact-head binding. Do not silently truncate requirements, accept invalid output, raise limits blindly or ask the consumer to split coherent work solely to obtain a green check.
  • Expose safe actual count/limit/phase facts and a concrete supported next action where available, without raw rejected provider output or secrets.
  • Separate this completed-result contract rejection from queue behavior, application test failures and optional check-publication errors.

This report requests upstream ownership only. Do not change the Domus consumer or its approved product scope as a workaround.

Activity

  1. added
    bugSomething isn't working
    bugfixSomething isn't working
    priority: highCritical issue, needs immediate attention.
    on Oct 31, 2026
  2. added theissue type on Oct 31, 2026
  3. moved this to Todo in vypdevon Oct 31, 2026
  4. changed the title [-]Progress scope assessment rejects an API requirement with scope-clause-limit in generated issue plan metadata[/-] [+]πŸ› - Progress scope assessment rejects an API requirement with scope clause limit in generated issue plan metadata[/+] on Oct 31, 2026
  5. vypbot commented on Oct 31, 2026

    @vypbot
    Contributor

    Hi! I’m @vypbot, the Copilot assistant for this repository.

    I can answer questions, explain the codebase, propose implementation and test plans, review issues and pull requests for potential bugs or security problems, and help authorized maintainers apply changes.

    Try /copilot help to see the available commands, or mention @vypbot with your question.

  6. efraespada commented on Oct 31, 2026

    @efraespada
    MemberAuthor

    Historical report 6541225162 is preserved verbatim in the immutable transcript. It remains incident evidence; the issue and SDD define acceptance. Original body SHA-256: 9ffeef2f783a488b45e3ac8962ce2e84418fb2e6e92c593f85fe86cca2a2a965.

  7. efraespada commented on Oct 31, 2026

    @efraespada
    MemberAuthor

    Historical report 6541808229 is preserved verbatim in the immutable transcript. It remains incident evidence; the issue and SDD define acceptance. Original body SHA-256: ad24d86eadc3ae3dbddd34dcd42510e1486f49fa962e6c42c520510b3821c39d.

  8. efraespada commented on Oct 31, 2026

    @efraespada
    MemberAuthor

    Historical report 6541906622 is preserved verbatim in the immutable transcript. It remains incident evidence; the issue and SDD define acceptance. Original body SHA-256: c12ea42f2f5093e4fb413f0cd896aef9fa5ee6c6a168e3df4130e9a2609da954.

  9. efraespada commented on Oct 31, 2026

    @efraespada
    MemberAuthor

    Historical report 6542265502 is preserved verbatim in the immutable transcript. It remains incident evidence; the issue and SDD define acceptance. Original body SHA-256: c98adc2fede78fe2e7fc439e6c4ebdec6b0731d4e005b30c402adf29fdbdd774.

  10. 26 remaining items

  11. vypbot commented on Nov 1, 2026

    @vypbot
    Contributor

    Issue closed β€” historical progress

    Current status: closed as completed.

    The last estimate is retained below as history. Closure does not certify a new percentage, tests, or branch alignment.

    Action required: none for this closed issue. If work continues, reopen the issue; a fresh assessment can update this same card.

    Last assessment before closure (historical)

    Progress: 55% β€” in progress

    > Current status: [R1] The reported source anchor is inside the Action-maintained hidden recommendationState.implementationPlan.steps details in the consumer issue body. That is an observation about the anchor, not proof that metadata inclusion, the scope bound or the model caused the failure. Please investigate which input and returned clause count exceeded which bound, whether derived recommendation metadata is intended to become progress scope, and the supported recovery for this coherent issue. Keep fail-closed validation and prior progress. β€” The patch adds host-side separation of valid persisted configuration from human scope, retains full-body bounds and freshness checks, and adds safe count, limit, phase, and source-line diagnostics. The reported incident's actual input and returned clause counts remain unverified; the patch does not establish its root cause.
    [R2] - Pin the original issue snapshot and distinguish human normative scope from generated or excluded metadata before choosing a fix. β€” The patch preserves physical source-line identities and separates a recognized persisted configuration block before human-line classification. It retains original full-body source bounds and snapshot freshness checks.
    [R3] - Exercise a legitimate API requirement with several nouns/invariants, long and short human descriptions, hidden generated recommendation JSON, repeated source anchors, and adversarial over-splitting. β€” Added regression cases cover coherent multi-
    [untrusted content truncated]

    Next: [R5] - If a bounded repair/retry is supported, preserve the complete scope, source identity, old progress and exact-head binding. Do not silently truncate requirements, accept invalid output, raise limits blindly or ask the consumer to split coherent work solely to obtain a green check. β€” Verify on the reported issue and exact comparison that the complete scope is accepted when within bounds, or rejected with prior progress retained when invalid or genuinely over limit.

  12. vypbot commented on Nov 1, 2026

    @vypbot
    Contributor

    Issue closed: synchronization recommendation retired

    No branch synchronization is required for this closed issue.

    The previous recommendation is historical. Issue closure does not certify that the branches are synchronized.

  13. moved this from In Progress to Done in vypdevon Nov 1, 2026
  14. added
    100%Progress: 100%
    and removed
    in-progressStart work on an admitted issue.
    state:workingWork can proceed on the verified branch or without a branch.
    on Nov 1, 2026
  15. efraespada commented on Nov 1, 2026

    @efraespada
    MemberAuthor

    Verified scope-clause-limit recurrence on Domus PR #15 at 6435decd55bfff32675d44e30f1d6a4e5d8eb161.

    The original job log states:

    Impact: The returned progress result violated its contract; progress was not published.
    Error code: agent.failed
    Details: phase scope, reason scope-clause-limit, field $.lines,
    requirement none, original issue line L111.
    Reference: 5c8115e7-6940-480a-8928-b5ac9c2bf49c
    

    I retrieved the current original issue body rather than guessing its line identity. L111 is exactly:

    |Household name/timezone|Explicit; backend 160/128 bounds and catalog|Draft, then authorized household command|
    

    The bounded error supplies no verified requirement ID, actual emitted clause array, measured clause count or allowed count. I therefore cannot attribute the overflow to a particular generated clause or infer a numeric limit. This is phase scope/$.lines, distinct from the preceding 119e504 phase assessment/$.requirements failure now reported in #829 (reference 1925868e-6b1f-4e09-950c-97abe93ea7df).

    No Copilot source, configuration, credentials, generated workflow, clause limit or verification rule changed. Domus continues with normal commits on the Action-managed branch. We have not shortened approved requirements, rewritten their original identity, started a rerun loop or used a replacement branch.

    Expected: generate a bounded schema-valid scope result that preserves the original requirement identities, or provide actionable measured facts for the offending output. Please resolve the generator/normalization contract in Copilot without requiring consumer acceptance or scope to be weakened.

  16. efraespada commented on Nov 1, 2026

    @efraespada
    MemberAuthor

    Another exact-head occurrence on consumer Domus PR15, source 7e80b1aa7ccc42a233f69c010d9dfa29323bfd4e.

    |\`GET /api/v1/operations/{operation_id}\`|Original own receipt plus actual pending/delivered/suppressed delivery, under current original-command authority|
    

    The previous head5415908 completed Commit and Verification successfully against the same unchanged original issue contract. The current increment only adds test/fixture counterevidence and documentation for the unrelated semantic Bugbot finding reported in#847; runtime and immutable Edge.4 build source are unchanged. The new scope failure points to L93 rather than the previous L111 occurrence. No raw generated clause count or numeric limit is supplied, so neither is inferred here.

    Existing progress labels/assessments were retained as the failure contract promises. Complete new-head project CI and Bugbot state are independent of this progress-generation failure. We continue implementation and preserve the owner requirement for complete product tests/coverage and no unresolved findings before merge. Copilot resources, limits, configuration, credentials and original issue acceptance have not been modified. Please own the upstream fix; no automatic rerun loop or weakened contract was introduced.

    Follow-up at ecdd85b4b08479eede41a517f9acdfdde9393b17: the same failure recurs in Commit and Verification, with phase scope, reason scope-clause-limit, field $.lines, requirement none, original issue line L103, reference 7cd0c149-da09-437a-9390-304e3e91d85a. Original L103 from the unchanged issue snapshot:

    Native retains actual Supervisor peer, accepted bounded prefix, current exact-human/binding checks and HA existing session; no second OAuth/Domus cookie. Technical Core authority never becomes a human role. Its nonpersonal shell/assets stay behind source/binding guards; personal APIs still require proof.
    

    This follow-up commit only explicitly closes a SQLite connection owned by a regression test; production/runtime/image source and the original issue contract remain unchanged. Full new-head CI is still separate. The semantic UI finding from#847 has meanwhile been natively resolved by vypbot as obsolete; this progress-generation failure is unrelated to that resolution. No limits, acceptance or Copilot files were changed.

    Follow-up at 6e05ba740af6c03cfff3509e0f687bac2a9f06f5: Commit job124925047062 and Verification124925466923 reject progress with phase scope, reason scope-clause-limit, field $.lines, requirement none, original issue line L97, reference 502f14a8-76d0-4e45-8aae-0bd7c73d4a4b. Existing progress is retained. No generated clause count/limit is supplied, so none is inferred. The independent project CI Check124926099504 succeeds, including actual Edge.5 images on both CPU families with19cases each. Native product findings are handled separately; the current documentation-version finding is being corrected. Copilot/source contract/credentials/gateway remain untouched; no retry loop or contract relaxation is used.

  17. efraespada commented on Nov 1, 2026

    @efraespada
    MemberAuthor

    Another observed scope-contract rejection on a new consumer issue, Domus #17, ready PR #19, head 930c2fc1f68e715f2c4ed39a2511fab1ae7fb235.

    This is reported against the same published immutable Action revision as the earlier observations. #828 is currently closed. I have not verified that a later fix is present in this published v3 artifact, and do not claim a regression on an unobserved newer revision. Please confirm the supported upstream recovery/publication or reopen/consolidate as appropriate.

    • Normal push comparison: ef49c3c093e9fce1fdc820dd945ab2c1da94b79e -> 930c2fc1f68e715f2c4ed39a2511fab1ae7fb235.
    • Commit run41344702698/job124961927315; Verification124962336982.
    • Original job resolves vypdev/copilot@v3 to 3e7ceea08e6131360e026937d1c0cc1b120472e0, runtime codex-cli0.166.0.
    • Agent admission20:02:57.192Z; scope rejection20:03:10.747Z on2026-11-01.
    • Phase scope, reason scope-clause-limit, field $.lines, requirement none, original issue source L162.
    • Reference 869f6f2e-5a73-4cdd-a78c-546d140d4d47; retryable Yes. Existing progress/assessments are retained and no valid new progress is published.

    The current original issue body was independently read. L162 is the human-reviewed prospective companion's state-table row:

    | Historical absent effect | Restored processed operation | It was processed, effect absent | Explicit new reviewed work | Never automatic recreation |
    

    This identifies the text present at inspection; the exact job-time body snapshot and rejected generated clause array are not available in the inspected diagnostics. Actual emitted clause count, configured bound and causal normalization behavior are therefore unknown. Unlike the initial issue description, this current anchor is a substantive human-supplied table row, not evidence of generated recommendation JSON inclusion.

    Independent exact-head CI Check succeeds, including both real CPU image gates. The new local temporal checkpoint passes3081 backend executions/110 authored modules at100% statement and branch coverage, with a source-complete audit. That does not establish accepted progress output or shipping the new temporal modules in the existing immutable image. The PR remains an incomplete implementation and is not being merged. Separate partial human-conversation Bugbot coverage is reported in#843 and is not this scope failure.

    No Copilot source, generated resource, credential, configuration, model, contract bound, consumer requirement or gateway was changed. No manual rerun loop was started. Please retain fail-closed output validation and prior progress while addressing generation/normalization or providing a source-preserving recovery; do not require truncating legitimate scope merely to obtain a green check.

  18. efraespada commented on Nov 1, 2026

    @efraespada
    MemberAuthor

    Exact-head follow-up on the same published progress-contract failure: Domus PR19, source 64908ee59705e6e9dc6e4a06a8d828bf5eea6d4f.

    Commit41346617406/job124967695376 and Verification124968110280 fail with phase scope, reason scope-clause-limit, field $.lines, requirement none, original issue line L173, reference bbc1ef0d-c919-444b-a9a9-6cf785935111. Existing progress is retained; generated clause count, bound and rejected output remain unavailable and are not inferred. The original issue body observed at inspection has L173: | Week start | Enum | MO | MO, TU, WE, TH, FR, SA, SU | Fixed rule, separate from display preference |.

    Independent exact-head CI and3260 backend tests/116 modules at100% pass. Current source was subsequently synchronized through the Action to580ab11; its new full CI and3398-test combined backend pass separately. These observations do not certify that the earlier rejected progress output became valid or that a newer upstream fix was deployed. The explicit new-head review's conversation limitation remains#843. No Copilot source/configuration/credentials, approved scope or bound changed; no manual retry loop was introduced.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

100%Progress: 100%branchedIndicates that a branch has been created for this issue.bugSomething isn't workingbugfixSomething isn't workingpriority: highCritical issue, needs immediate attention.size: XXLIndicates an extremely large task or issue that requires significant time and effort to complete.

Type

Projects

  • Status
    Done

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions