docs(specs): #85 bug prebuild — canary PR body in reviewer context #363

Merged
pdurlej merged 1 commit from claude/fork-C1-canary-pr-body-context-bug into main 2026-05-23 09:19:09 +02:00
Collaborator

Summary

Fork C1 retry (2026-05-18) of work lost in 2026-05-17 fork-anchor collision. Per dispatch template v0.1 (PR #356), worktree-isolated at /Users/pd/Developer/iskra-platform-2026-04-30-fork-C1.

Bug shape (per Codex 2026-05-17 feedback) — reproduction + hypothesis ranking + minimal fix surface + acceptance + tests. Not feature Spec Kit.

Smoking gun (root cause CONFIRMED via static trace)

  • .forgejo/workflows/canary-required.yml:190-201 invokes run_review.py without --pr-description-file flag
  • run_review.py:405-407 defaults pr_description to ""
  • control-plane/platformctl/tools/providers/_base.py:199 renders "Description:\n(none)" on empty input
  • Reviewers see "PR description: none" → false dissent → 3-iter cap on PR #84

Receiver + provider are correct. Bug is single-caller-side.

Files

  • docs/specs/canary-pr-body-context-bug-prebuild/README.md — overview + safety boundary
  • docs/specs/canary-pr-body-context-bug-prebuild/01-reproduction.md — repro + evidence from PRs #82 #84
  • docs/specs/canary-pr-body-context-bug-prebuild/02-hypothesis-ranking.md — 6 hypotheses ranked; #1 confirmed
  • docs/specs/canary-pr-body-context-bug-prebuild/03-minimal-fix-surface.md — Path A (workflow only) vs Path B (workflow + receiver pr_description_status); recommend B
  • docs/specs/canary-pr-body-context-bug-prebuild/04-acceptance.md — AC-1..AC-6 + regression fixtures + anti-acceptance
  • prompts/codex-canary-pr-body-context-bug.md — companion execution prompt

Path B recommendation

Path A fixes the bug. Path B makes future regressions of this fix structurally distinct in agent-readable signal — future cousins reading reviewer output see "(body fetch was not requested by caller — possible regression of #85)" instead of ambiguous (none). Per ADR-0018: agentic-first systems pay the cost of "signal-clean state" once, not at every future investigation.

Cost delta: ~40 LoC additional vs Path A. Still Lite tier.

Recursive validation

Implementation PR's own canary run is the smoke. If the fix works, the PR landing the fix will see its OWN body in its OWN reviewer prompts (no "description: none" dissent). Definitive proof without operating on production traffic.

Safety / production boundary

This PR prepares implementation only. It does NOT authorize runtime mutation, production restart, credential changes, sacred-path modification, or any platformctl apply invocation.

Implementation slice lands as a separate Lite-tier PR (likely codex/fix/85-canary-pr-body-context) following canary 3+3 + operator merge per ADR-0007 + ADR-0017 + ADR-0018.

Tier

Trivial per ADR-0007 (docs + execution prompt only; no code/schema/runtime/sacred-path mutation in THIS PR).

Why per ADR-0018

The existing situation (canary reviewers cap-out due to false description: none dissent) is exactly the misleading-signal failure mode ADR-0018 was written to prevent. Cousins reading PR #84's canary output saw high-dissent terminal state and could not distinguish "real review failure" from "input pipeline missing the description". Single-location root-cause fix per ADR-0018.

Refs #85 #82 #84

## Summary Fork C1 retry (2026-05-18) of work lost in 2026-05-17 fork-anchor collision. Per dispatch template v0.1 (PR #356), worktree-isolated at `/Users/pd/Developer/iskra-platform-2026-04-30-fork-C1`. **Bug shape** (per Codex 2026-05-17 feedback) — reproduction + hypothesis ranking + minimal fix surface + acceptance + tests. Not feature Spec Kit. ## Smoking gun (root cause CONFIRMED via static trace) - `.forgejo/workflows/canary-required.yml:190-201` invokes `run_review.py` **without `--pr-description-file` flag** - `run_review.py:405-407` defaults `pr_description` to `""` - `control-plane/platformctl/tools/providers/_base.py:199` renders `"Description:\n(none)"` on empty input - Reviewers see `"PR description: none"` → false dissent → 3-iter cap on PR #84 Receiver + provider are correct. Bug is **single-caller-side**. ## Files - `docs/specs/canary-pr-body-context-bug-prebuild/README.md` — overview + safety boundary - `docs/specs/canary-pr-body-context-bug-prebuild/01-reproduction.md` — repro + evidence from PRs #82 #84 - `docs/specs/canary-pr-body-context-bug-prebuild/02-hypothesis-ranking.md` — 6 hypotheses ranked; #1 confirmed - `docs/specs/canary-pr-body-context-bug-prebuild/03-minimal-fix-surface.md` — Path A (workflow only) vs Path B (workflow + receiver `pr_description_status`); recommend B - `docs/specs/canary-pr-body-context-bug-prebuild/04-acceptance.md` — AC-1..AC-6 + regression fixtures + anti-acceptance - `prompts/codex-canary-pr-body-context-bug.md` — companion execution prompt ## Path B recommendation Path A fixes the bug. **Path B makes future regressions of this fix structurally distinct in agent-readable signal** — future cousins reading reviewer output see `"(body fetch was not requested by caller — possible regression of #85)"` instead of ambiguous `(none)`. Per ADR-0018: agentic-first systems pay the cost of "signal-clean state" once, not at every future investigation. Cost delta: ~40 LoC additional vs Path A. Still Lite tier. ## Recursive validation Implementation PR's own canary run is the smoke. If the fix works, the PR landing the fix will see its OWN body in its OWN reviewer prompts (no `"description: none"` dissent). Definitive proof without operating on production traffic. ## Safety / production boundary **This PR prepares implementation only.** It does NOT authorize runtime mutation, production restart, credential changes, sacred-path modification, or any `platformctl apply` invocation. Implementation slice lands as a separate Lite-tier PR (likely `codex/fix/85-canary-pr-body-context`) following canary 3+3 + operator merge per ADR-0007 + ADR-0017 + ADR-0018. ## Tier **Trivial** per ADR-0007 (docs + execution prompt only; no code/schema/runtime/sacred-path mutation in THIS PR). ## Why per ADR-0018 The existing situation (canary reviewers cap-out due to false `description: none` dissent) is exactly the misleading-signal failure mode ADR-0018 was written to prevent. Cousins reading PR #84's canary output saw high-dissent terminal state and could not distinguish "real review failure" from "input pipeline missing the description". Single-location root-cause fix per ADR-0018. Refs #85 #82 #84
docs(specs): #85 bug prebuild — canary PR body in reviewer context
All checks were successful
base-is-main / guard (pull_request) Successful in 1s
canary-required / collect-diff (pull_request) Successful in 4s
patchwarden-pr-sanity / collect-diff (pull_request) Successful in 4s
canary-required / canary (pull_request) Successful in 13s
patchwarden-pr-sanity / sanity (pull_request) Successful in 21s
dc53f6e213
Fork C1 retry (2026-05-18) of work lost in 2026-05-17 fork-anchor
collision. Per dispatch template v0.1 (PR #356), worktree-isolated.

Bug shape (per Codex 2026-05-17 feedback) — NOT feature Spec Kit:
- README.md (overview + safety boundary)
- 01-reproduction.md (steps + evidence from PRs #82, #84)
- 02-hypothesis-ranking.md (6 hypotheses ranked; #1 confirmed
  via static trace)
- 03-minimal-fix-surface.md (Path A: workflow YAML only;
  Path B: + receiver-side pr_description_status — recommended)
- 04-acceptance.md (AC-1..AC-6 + regression fixtures + anti-acceptance)

Plus companion execution prompt:
- prompts/codex-canary-pr-body-context-bug.md

Smoking gun found by static trace:
- .forgejo/workflows/canary-required.yml:190-201 invokes
  run_review.py without --pr-description-file flag
- run_review.py:405-407 defaults pr_description to ""
- _base.py:199 renders "Description:\n(none)" on empty input
- Reviewers see "PR description: none" → false dissent → 3-iter cap
  on PR #84

Bug is single-caller-side. Receiver + provider are correct.
Path B fix makes future regressions structurally distinct in
agent-readable signal (per ADR-0018: future cousins must read
green as green).

Tier: Trivial per ADR-0007 (prebuild = docs + execution prompt;
no code/schema/runtime/sacred-path mutation in THIS PR).
Implementation slice lands as separate Lite PR.

This PR prepares implementation only. It does NOT authorize
runtime mutation, production restart, credential changes, or cleanup.

Refs #85 #82 #84
Sign in to join this conversation.
No reviewers
No labels
W6d-automerge-calibration
agent/claude-code
agent/codex
agent/hermes
agent/iskra
agent/ollama
agent/patchwarden
automerge-candidate
class/security-sensitive
cutover-gate
dependency/blocked
dependency/blocks-others
dependency/cross-repo
dependency/needs-confirmation
domain:agents
domain:ci
domain:docs
domain:forgejo
domain:infra
domain:memory
domain:runtime
domain:signal
domain:ux
flow/architecture
flow/blocked
flow/deployed
flow/done
flow/implementation
flow/intake
flow/maintained
flow/observed
flow/ready
flow/refining
flow/retired
flow/review
iterating
judge/codex-candidate
judge/hermes-candidate
judge/low-confidence
judge/needs-refinement
judge/operator-needed
judge/p0
judge/p1
judge/p2
judge/p3
judge/park
judge/patchwarden-candidate
judge/stale-priority
kind/adr
kind/bug
kind/chore
kind/feature
kind/infra
kind/ops
kind/refactor
kind/research
large-impact
merge/auto
merge/manual
merge/manual-dependency-conflict
merge/manual-failing-tests
merge/manual-merge-conflict
merge/manual-missing-review
merge/manual-operator-preference
merge/manual-red-zone
merge/manual-security-sensitive
merge/manual-unclear-scope
merge/manual-unknown
meta
mode:operator-only
mode:patchwarden-iskra-approved
mode:safe-auto
needs-operator-decision
needs-triage
not-ready
observed/erroring
observed/needs-followup
observed/pending
observed/retire-candidate
observed/unused
observed/used
operator-emotional
owner-attention
phase/02
phase/03
priority:p0
priority:p1
priority:p2
priority:p3
proposed
ready-for-agent
ready-for-operator
recovery
review:claude-reviewed
review:codex-reviewed
review:dziadek-reviewed
review:needs-human
risk/exposure
risk/process
risk/product
risk/runtime
safety:external-write
safety:no-prod-mutation
safety:prod-impact
safety:secret-touch
size/large
size/medium
size/small
size/tiny
size/unknown
source/adr
source/agent-generated
source/manual
source/operator-chat
source/voice-note
status:blocked
status:codex-ready
status:merged:pending-evidence
status:needs-evidence
status:operator-needed
status:parked
tier/full
tier/lite
tier/stacked
tier:0-platform-substrate
tier:1-iskra-value-layer
tier:2-tools-products-modules
type:bug
type:chore
type:docs
type:feat
type:policy
type:research
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
pdurlej/platform!363
No description provided.