docs: record the 2026-09-02 product-goal-directive revisions (2nd + 3rd) - #1692
docs: record the 2026-09-02 product-goal-directive revisions (2nd + 3rd)#1692seonghobae wants to merge 11 commits into
Conversation
The owner reissued the full nine-section autonomous PR/merge/development loop directive with substantially expanded text (core/consumer boundary policy in §2, i18n and Keyverse-scoping detail in §4, ontology-ownership split and broadened identifier-naming rule in §5, an explicit orchestrator/free pin folded into §8, and a full named core-repo table with a core-vs-domain-product classification in §9). Recorded verbatim per this file's own conflict policy. Verified all 29 repos named in §5/§9 exist under ContextualWisdomLab with exact-matching case, and reconciled the new §5 identifier-naming rule against CWL-MASTER-CONTEXT.md §7's DB-object grandfather clause (still binding, now with an explicit note that the broader rule applies going forward on touched/new code, not as a mandate to force-rename existing identifiers ecosystem-wide). Flagged, but did not resolve, an open tension between §9's quarantine-sandbox-runtime repo and CWL-MASTER-CONTEXT.md §3's description of noema owning the sandbox — needs each repo's own docs read before deciding, not guessed. Nine-section transcription independently verified byte-for-byte against the source text by a separate review pass before this commit. See docs/doctoring/product-goal-directive.md for the full reconciliation record. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016ErwZSYW3pm585NiM3Q7aN
|
Warning Review limit reachedNext included review available in 16 minutes. View limit detailsLimit details: You’ve used the included review currently available. You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Team Run ID: 📒 Files selected for processing (4)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
The owner issued one further refinement of the directive after the second revision (already on this branch/PR): §1 clarifies "0 open PRs" must come from merges or full successor-absorption, never a bare close; §2 gains a substantial repair-vs-close PR-lifecycle policy; §4 drops the named-framework shortlist for framework-agnostic criteria; §5 moves its per-repo ontology breakdown into §9; §8 spells out exactly how the orchestrator/free pin must be implemented at the workflow layer; §9 adds a definition of what "core foundation" means and reorganizes the repo table into five categories with concrete interim-boundary mechanisms. Continuing the existing branch/PR per the directive's own "edit this file in place, do not fork a second copy" policy, applied to the PR level too, since #1692 was still open when this third revision arrived. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016ErwZSYW3pm585NiM3Q7aN
Three confirmed findings from Devin Review's pass on PR #1692, fixed in place: - docs/product-goal-directive.md §5's rename-reconciliation note told an agent to "pause and confirm" before a contract-breaking rename, which directly contradicts docs/CWL-MASTER-CONTEXT.md §7's binding "Do NOT ask the user to decide" full-autonomy convention. Replaced with an autonomous, contract-safe process (preserve the published boundary, prepare a versioned migration plan, record an ADR) instead of asking. - The repo-count arithmetic in docs/product-goal-directive.md §9 and its doctoring record was wrong: §9 actually lists 23 core repos (not 21), and the four ontology repos named in §5 are a named subset of those 23, not an additional group — so the correct total is 23 + 7 domain-consumer repos = 30 unique names, not 29. Fixed both files' counts and the double-counting framing. - The §9 "sandbox ownership" reconciliation item was left flagged as open rather than resolved. Read quarantine-sandbox-runtime's and noema's actual current READMEs (plus noema's own sandbox-planning doc, which states the review agent runs in "a separate quarantined execution plane" and must not run untrusted code in the Noema Worker process): quarantine-sandbox-runtime is the current AI-SOC sandbox owner, not noema. Updated docs/CWL-MASTER-CONTEXT.md at all four locations that said "noema quarantine sandbox" (§3, §6 header, the P1 roadmap bullet, and the ecosystem UML diagram's node/edge) to name quarantine-sandbox-runtime instead, and marked the directive's reconciliation note resolved. docs/doctoring/product-goal-directive.md records the full reasoning for all six Devin findings (three fixed, one requiring cross-repo evidence, two informational/no-action) per this file's existing pattern from the original PR #1429 Devin Review pass. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016ErwZSYW3pm585NiM3Q7aN
…ring section
Devin Review caught a residual inconsistency the previous fix missed: the
doctoring file's third-revision section ("What changed in the directive
text" and its own "Verification performed this revision") repeated the
same stale "21 core repos" figure twice, independently of the
second-revision section already corrected to 23. Fixes both remaining
occurrences.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ErwZSYW3pm585NiM3Q7aN
|
Fresh owner-stack reconciliation before write: protected I will preserve the exact head tree and both histories with a non-force two-parent reconciliation against current protected main. This is ancestry repair only; it does not promote a Proposed/open owner contract to protected authority, transfer old Checks, or discard concurrent main intent. A fast-forward ref update will be attempted only if the branch head is still exactly |
…ective-2026-09-02-revision
Current purpose — 2026-09-02
This PR is the single writer for the standing
docs/product-goal-directive.mdrevision, its doctoring record, and the directly conflictingdocs/CWL-MASTER-CONTEXT.mdarchitecture boundary. It carries CWL DEVELOPMENT PHILOSOPHY v2026-09-02B without creating a competing directive copy.Current exact head:
e7c5eec969a4a063769f53c89e130f62d18bef6aondocs/product-goal-directive-2026-09-02-revision.Current-main reconciliation
Protected
.githubmainis78271917b526469c559fa75cb5ee39426e5494d1. The writer is now non-destructively reconciled with that exact protected head (behind_by=0). A first reconciliation accidentally reused the stale branch tree and temporarily widened the PR to 26 files; that defect was immediately repaired by rebuilding the tree from protected main and overlaying only owner-path blobs. No force push, destructive rebase, branch reset, or concurrent protected-main delta deletion was used.The current effective delta is exactly four files:
docs/product-goal-directive.mddocs/doctoring/product-goal-directive.mddocs/CWL-MASTER-CONTEXT.mdtests/test_product_technical_gap_baseline.pyQuarantine ownership RCA and contract
The product directive already identified
quarantine-sandbox-runtimeas the isolation owner, while the master context still contained residual claims that Noema owned plugin quarantine and was the shared quarantine runtime. Direct owner evidence already recorded in this PR establishes the opposite boundary: Noema is the GitHub Actions OIDC short-lived repository-capability / exact-revision evidence control plane;quarantine-sandbox-runtimeowns untrusted artifact/code isolation.A genuine RED regression commit
7deae853bbb7697dab79c235b21340ba03f5f7ea, based on pre-fix source61104eabac120e3751b1daec399538ff0d689fb9, rejects the two residual Noema-quarantine phrases and requires both the canonical Noema responsibility and UML edges. The production source repair is575f5a0c71048e837b62d64de92ec9dedd56e9be; merge successore7c5eec969a4a063769f53c89e130f62d18bef6aintegrates RED + repair without force. Exact-head source inspection confirms the forbidden current-role phrases are absent and the canonical owner markers are present.Admission state
All existing review threads are resolved on the current diff. Fresh exact-head Python Security, Secret Scan, Scorecard, OSV, Semgrep, CodeQL, Security Scan, and SBOM runs have been regenerated and are queued/pending; predecessor results do not transfer. This governance PR remains Proposed / not merge-ready until exact-current-head required evidence is terminal and any new substantive finding is repaired. Do not administrator-bypass it solely because the central Actions fleet is saturated.