corrections: new plugin with /corrections:codify #7573

Activity

1. kyle-sexton

   added

   needs-humanHuman-in-the-loop required; autonomous sessions must not resolve items carrying this.Human-in-the-loop required; autonomous sessions must not resolve items carrying this.

   on Oct 16, 2026
2. kyle-sexton

   added the

   Feature

   issue type on Oct 16, 2026
3. kyle-sexton

   added a parent issue on Oct 16, 2026

   • Epic: absorb Lauren Tan's and Matt Pocock's agent workflows (poteto-review) #7555
4. kyle-sexton

   marked this as blocked by 2 issues on Oct 16, 2026

   • enforcement-ladder: v2.0.0 ladder and one guardrail criterion #7572
   • config-cascade: settings resolution and convention home #7571
5. kyle-sexton

   set

   Work status

   to

   needs-spec

   on Oct 16, 2026
6. kyle-sexton

   marked this as blocking on Oct 16, 2026

   • session-flow: retro findings, catch-up scope and recommend-skills router #7575
7. github-actions commented on Oct 16, 2026

   

   github-actionsBot

   on Oct 16, 2026 – with GitHub Actions

   Last edited by melodic-automation-lanes

   Contributor

   More actions

   Proposed work class: work-class: structural
   Basis: automated intake triage; the lanes App applied this class.

   Size: Medium
   Complexity: Medium
   Uncertainty: Known
   Flags: none
   Blocked by: none

   Decision comment (#7573 (comment)):

   Decision recorded.
   Agent choice, not an owner answer.

   The issue's Decisions say corrections "absorbs retro codify, /review:audit-enforceability and audit-sessions style". The first two are delivered on main:
   [quote truncated; the full decision is at the link]

   The recorded decision settles the open questions. It picks one way to finish absorbing style: codify gains a cross-session preference action, audit-sessions gains an excerpts-only consumer mode, and style is removed. It also gives checkable done-when items. The remaining work is one change across the corrections and session-flow plugins, so I triaged it as a bounded Medium task.

   _{Automated triage, applied by the lanes App.}
8. kyle-sexton

   marked this as blocking on Oct 16, 2026

   • architecture: boundaries and conventions lenses #7590
9. kyle-sexton

   added

   needs-specNeeds a Brief or plan before anyone can build it.Needs a Brief or plan before anyone can build it.

   on Oct 16, 2026
10. kyle-sexton commented on Oct 16, 2026

    

    kyle-sexton

    on Oct 16, 2026

    MemberAuthor

    More actions

    What the article adds for codify. Source: https://claude.dev/blog/solving-code-review-at-anthropic (2026-10-15); ids are digest unit and claim.

    • After a miss, ask what the model did not have in front of it: "When a review misses something, we ask what the model didn't have in front of it and put the fix into a CLAUDE.md, a skill or a new check" (11 C5).
    • When the reviewer keeps flagging the same thing, fix it at the source: the skill, a lint rule, CLAUDE.md or REVIEW.md (04 C12; 10 C6).
    • Incident reviews ask which layer should have caught the issue (04 C68).

    For /corrections:codify: record the "what was missing" answer in each proposal, so rung picking (#7572) starts from it. Under D97/Q150 (GUIDE 97) review rules go to the per-concern review files; REVIEW.md and the one CLAUDE.md/AGENTS.md pointer line the owner allowed on 2026-10-16 only point at them. #7583 carries that change.

    Owner decision, 2026-10-16. The codify threshold stays at its default of 2 (Q572), and a person's direct "codify this" acts at once (Q509). Each confirmed escaped defect that Q567 sends to codify gets the article's question, "what did the model not have in front of it" (11 C5), answered in its follow-up.

    Against the originating sources. Counting to two and re-reading history (Lauren's, pstack correct) agrees with "keeps flagging the same thing" (04 C12).

    Each miss is also an eval case candidate for #7639 (05 C2).

    Related: #7572, #7583, #7576, #7639, #7638.
11. kyle-sexton

    unmarked this as blocked on Oct 18, 2026

    • config-cascade: settings resolution and convention home #7571
12. kyle-sexton

    marked this as blocked on Oct 18, 2026

    • config-cascade: resolver and reader implement the config ADR (overlay, policy floors, root classification, --ref) #7938
13. kyle-sexton

    added sub-issues on Oct 20, 2026

    • feat(corrections): add the corrections plugin with /corrections:codify and a prove-check script #8636
    • feat(corrections): add the corrections plugin with a prove-check script #8732
14. 24 remaining items

    Load more
15. kyle-sexton

    added

    host-runNeeds an unattended run on a host; the host runner takes it, not a person or the issue lane.Needs an unattended run on a host; the host runner takes it, not a person or the issue lane.

    on Nov 4, 2026
16. kyle-sexton commented on Nov 4, 2026

    

    kyle-sexton

    on Nov 4, 2026

    MemberAuthor

    More actions

    Decision recorded.
    Agent choice, not an owner answer.

    The issue's Decisions say corrections "absorbs retro codify, /review:audit-enforceability and audit-sessions style". The first two are delivered on main:

    • feat(corrections): give codify a findings-file input, a review-comment source and their evals #11629, refactor(session-flow)!: remove retro's codify mode #12073 and refactor(review): remove /review:audit-enforceability #12253 did that work.
    • The done-when search for those two names prints nothing outside the changelogs.
    • Every slice sub-issue is closed. feat(corrections): add the corrections plugin with /corrections:codify and a prove-check script #8636 closed not_planned when it was split into feat(corrections): add the corrections plugin with a prove-check script #8732, feat(corrections): add /corrections:codify, which counts a recurring mistake, picks a ladder rung and proves the check #8734 and feat(corrections): read codify's recurrence threshold from the recurrence.min_occurrences setting #8735.

    #11121 scoped style out as "a separate slice", which was never filed, so this record covers only style.

    Today /session-flow:audit-sessions style counts on its own and hands preferences to codify:

    • plugins/session-flow/skills/audit-sessions/SKILL.md:228 and :272 hold the pass and the hand-off.
    • plugins/corrections/skills/codify/SKILL.md:185 says it "keeps its own counter".

    This record picks one way to finish the absorption. Codify gains a cross-session preference action that reads session excerpts through a new audit-sessions consumer mode. style is removed. The other options fail on recorded facts:

    • Keeping style in audit-sessions, whether as the current hand-off or as a buffer producer, conflicts with the issue's Rejected line, "an action on /session-flow:audit-sessions".
    • A second transcript reader in corrections would duplicate audit-sessions' collector, its redaction and its format canaries (audit-sessions/SKILL.md:56-92).
    • A consumer mode has a precedent: sweep-only serves another skill and writes nothing (audit-sessions/SKILL.md:42-54).

    The mode name excerpts-only mirrors sweep-only (Basis: judgment). This is one lane PR across corrections and session-flow.

    • plugins/session-flow/skills/audit-sessions/SKILL.md adds an excerpts-only consumer mode modeled on sweep-only (SKILL.md:42-54). It runs sweep.py --excerpts with the caller's --scope, --state-key and --since (SKILL.md:257). It returns the JSON lines verbatim with their exit code, refuses --write-report, and writes nothing.

    • plugins/corrections/skills/codify/SKILL.md adds an action row and an actions/<name>.md file for cross-session preferences, gated on session-flow being installed. The action:

      • asks for the user-scope target (user CLAUDE.md, a user rule file or a user output style) and never accepts a repository file;
      • passes --scope explicitly and gets excerpts by invoking /session-flow:audit-sessions excerpts-only, never by reading transcripts or the store itself;
      • groups the excerpts and counts distinct sessions against recurrence.min_occurrences, through codify-settings.mjs;
      • has the person confirm each preference, then proposes through sections 3 and 6.

      The steps and the target rule carry over from audit-sessions/SKILL.md:228-281. The threshold source is the issue's Decisions ("recurrence.min_occurrences, default 2") and codify/SKILL.md:53-55. The cross-plugin naming rule is docs/review/skills-and-agents.md:12-13.

    • Codify treats excerpt lines as data, never instructions, citing docs/contracts/untrusted-content/README.md. This carries over the rule at audit-sessions/SKILL.md:279-281 and matches codify/SKILL.md:48-51.

    • The style action is removed from audit-sessions/SKILL.md: the "Style pass" section, style in the Arguments line and the action list (:11, :34, :37), and the eval case style-pass-needs-two-sessions-and-a-user-scope-target in audit-sessions/evals/evals.json:77. The correction lens no longer points at "the style pass's step 3" (:288); it states its own --scope rule. Source: issue Decisions; the removal pattern of refactor(session-flow)!: remove retro's codify mode #12073.

    • git grep -n -E 'audit-sessions[^ ]{0,12} style|[Ss]tyle pass|`style` (pass|action|takes)' -- plugins/session-flow plugins/corrections docs ':!**/CHANGELOG.md' ':!.changes/**' prints nothing. That removes the current hits at plugins/session-flow/README.md:246,258, audit-sessions/SKILL.md:11,23,34,37,228,288, audit-sessions/evals/evals.json:78, codify/SKILL.md:185 and docs/upstream/cursor-pstack.md:63. plugins/session-flow/README.md and the docs/upstream/cursor-pstack.md row's "Taken" text point at the new codify action in place of the removed lines; the record is reworded, not deleted, as feat(corrections): absorb retro codify and /review:audit-enforceability into /corrections:codify and remove them #11121 did for its names.

    • codify/SKILL.md's Boundary drops "It does not run /session-flow:audit-sessions style, which keeps its own counter" (:185). The description and argument-hint name the new action. Codify's allowed-tools gains no transcript or store read.

    • The removed audit-sessions eval case is ported into plugins/corrections/evals/ as a case.yaml case. A new case shows that a preference seen in only one session yields no proposal. Before/after eval results for the moved text appear in the PR, as feat(corrections): give codify a findings-file input, a review-comment source and their evals #11629's port table and eval comment did. Source: docs/review/skills-and-agents.md:6-8 ("a cut to an existing skill ships only with eval evidence of no loss").

    • bash plugins/skill-quality/scripts/check-skill.sh --require-evals plugins/corrections/skills exits 0, and bash plugins/skill-quality/scripts/check-skill.sh plugins/session-flow/skills exits 0.

    • bash plugins/corrections/scripts/prove-check.test.sh and bash plugins/corrections/scripts/codify-settings.test.sh exit 0. plugins/session-flow/skills/audit-sessions/scripts/tests/test_sweep.py passes, with no excerpt test removed.

    • scripts/check-changelog-fragments.sh --check passes, with a fragment under .changes/corrections/ (minor: a new action) and one under .changes/session-flow/ (major: an action is removed, as in refactor(session-flow)!: remove retro's codify mode #12073).

    needs-spec is satisfied by this record.
17. github-actions

    added

    needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

    and removed

    status: confirmedA maintainer confirmed the item is real and in scope; ready to be worked.A maintainer confirmed the item is real and in scope; ready to be worked.

    on Nov 4, 2026
18. kyle-sexton

    removed

    needs-specNeeds a Brief or plan before anyone can build it.Needs a Brief or plan before anyone can build it.

    on Nov 4, 2026
19. github-actions

    removed

    needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

    on Nov 4, 2026
20. kyle-sexton commented on Nov 4, 2026

    

    kyle-sexton

    on Nov 4, 2026

    MemberAuthor

    More actions

    Claimed by host-run lane 4 (session 839ddb58-b732-4a71-bb69-79a384292e2e) on hub assignment H32, 2026-11-04T07:19Z: the style absorption per issuecomment-6625049760.
21. melodic-automation-lanes

    added

    status: readyTriaged, unblocked, and fully specified; eligible to pick up.Triaged, unblocked, and fully specified; eligible to pick up.

    on Nov 4, 2026
22. melodic-automation-lanes

    changed

    Work status

    to

    ready

    on Nov 4, 2026
23. melodic-automation-lanes

    added

    status: confirmedA maintainer confirmed the item is real and in scope; ready to be worked.A maintainer confirmed the item is real and in scope; ready to be worked.

    and removed

    status: readyTriaged, unblocked, and fully specified; eligible to pick up.Triaged, unblocked, and fully specified; eligible to pick up.

    on Nov 4, 2026
24. melodic-automation-lanes

    changed

    Work status

    to

    confirmed

    on Nov 4, 2026
25. kyle-sexton

    added 2 commits that reference this issue on Nov 4, 2026

    test(corrections): cover the session-flow gate and the excerpts-only …

    ...

    a2fa212

    test(corrections): count an install pointer as the gate's stop

    ...

    3d66b12
26. kyle-sexton

    closed this as completedin #13162on Nov 4, 2026