feat(agent-waves): pilot OpenAI Saver wave #746

Merged
pdurlej merged 1 commit from codex/openai-saver-pilot into main 2026-06-08 15:03:39 +02:00
Collaborator

Canary status: missing — fire canary 3+3 manually before merge

Canary Context Pack

Product story

OpenAI weekly budget is tight while Ollama and AntiGravity/Gemini capacity is mostly unused. This PR introduces the smallest safe OpenAI Saver pilot: deterministic cloud-dispatch protection first, then artifact-first model scouting that gives Codex a bounded packet instead of broad exploration.

What changed

  • Patchwarden PR sanity now skips Ollama Cloud model dispatch when deterministic findings contain secret/trust-boundary blockers.
  • Added focused tests proving secret/workflow blockers skip cloud review and safe docs diffs still dispatch review.
  • Added reusable agent-wave template, AntiGravity/agy prompts, and a completed sanitized deslop wave for #726-731.

Why it changed

Cloud reviewers are useful only if unsafe diffs are stopped before egress. The wave artifacts prove the workflow on current deslop issues and select #726 as the next executable Codex slice.

Files touched

  • control-plane/platformctl/ci/pr_sanity.py
  • control-plane/platformctl/tests/test_pr_sanity.py
  • docs/agent-waves/antigravity-prompts.md
  • state/agent-wave-template.md
  • state/waves/2026-06-08-openai-saver-deslop/

Relevant context

  • docs/forgejo-agent-operations.md Ollama Cloud review rules
  • decisions/0023-gemini-3-5-flash-reviewer-cousin.md
  • state/cycle/W6d-autonomous-merge-pilot.md
  • Oracle consult summarized in-thread before this PR

Runtime evidence

N/A — no runtime mutation. Read-only Forgejo issue inspection only.

Known constraints

  • Model outputs are evidence, not approval or merge authority.
  • #731 remains owner-gated runtime work.
  • #729 remains security-sensitive workflow/CI work.
  • Kimi timed out; the timeout is recorded as a coverage gap, not approval.

Explicit out-of-scope

  • Runtime deploy/restart/apply/migration.
  • Forgejo issue/PR mutation by scout models.
  • Implementing #726 in this PR.
  • Merging or automerge changes.

Requested decision

approve_merge after full review/canary, with class/security-sensitive label.

Merge blockers

  • Any concern that secret/trust-boundary diffs could still reach cloud model review.
  • Invalid JSON artifacts.
  • Failed test_pr_sanity.py focused test suite.

Spec sources read

  • control-plane/platformctl/ci/pr_sanity.py: model-review dispatch and deterministic findings.
  • control-plane/platformctl/tests/test_pr_sanity.py: existing PR sanity coverage and added regression tests.
  • docs/forgejo-agent-operations.md: Ollama Cloud review skip rules and Patchwarden boundary.
  • decisions/0023-gemini-3-5-flash-reviewer-cousin.md: Gemini/AntiGravity advisory lane.
  • state/cycle/W6d-autonomous-merge-pilot.md: Patchwarden model split and merge boundary.
  • Forgejo issues #726-#731 via read-only API: wave target metadata and issue bodies.

Verification

PYTHONPATH=control-plane uv run --project control-plane --extra dev python -m pytest control-plane/platformctl/tests/test_pr_sanity.py
35 passed

python3 -m json.tool <wave json files>
json ok

rg secret-pattern scan over docs/agent-waves, state/agent-wave-template.md, and wave artifacts
no matches

git diff --check
clean

Test plan

  • Confirm deterministic secret/trust-boundary blockers produce skipped model reviews.
  • Confirm safe docs diffs still dispatch configured model reviews.
  • Confirm wave JSON artifacts are valid and sanitized.
  • Run full security-sensitive review/canary before merge.
Canary status: missing — fire canary 3+3 manually before merge ## Canary Context Pack ### Product story OpenAI weekly budget is tight while Ollama and AntiGravity/Gemini capacity is mostly unused. This PR introduces the smallest safe OpenAI Saver pilot: deterministic cloud-dispatch protection first, then artifact-first model scouting that gives Codex a bounded packet instead of broad exploration. ### What changed - Patchwarden PR sanity now skips Ollama Cloud model dispatch when deterministic findings contain secret/trust-boundary blockers. - Added focused tests proving secret/workflow blockers skip cloud review and safe docs diffs still dispatch review. - Added reusable agent-wave template, AntiGravity/agy prompts, and a completed sanitized deslop wave for #726-731. ### Why it changed Cloud reviewers are useful only if unsafe diffs are stopped before egress. The wave artifacts prove the workflow on current deslop issues and select #726 as the next executable Codex slice. ### Files touched - `control-plane/platformctl/ci/pr_sanity.py` - `control-plane/platformctl/tests/test_pr_sanity.py` - `docs/agent-waves/antigravity-prompts.md` - `state/agent-wave-template.md` - `state/waves/2026-06-08-openai-saver-deslop/` ### Relevant context - `docs/forgejo-agent-operations.md` Ollama Cloud review rules - `decisions/0023-gemini-3-5-flash-reviewer-cousin.md` - `state/cycle/W6d-autonomous-merge-pilot.md` - Oracle consult summarized in-thread before this PR ### Runtime evidence N/A — no runtime mutation. Read-only Forgejo issue inspection only. ### Known constraints - Model outputs are evidence, not approval or merge authority. - `#731` remains owner-gated runtime work. - `#729` remains security-sensitive workflow/CI work. - Kimi timed out; the timeout is recorded as a coverage gap, not approval. ### Explicit out-of-scope - Runtime deploy/restart/apply/migration. - Forgejo issue/PR mutation by scout models. - Implementing #726 in this PR. - Merging or automerge changes. ### Requested decision `approve_merge` after full review/canary, with `class/security-sensitive` label. ### Merge blockers - Any concern that secret/trust-boundary diffs could still reach cloud model review. - Invalid JSON artifacts. - Failed `test_pr_sanity.py` focused test suite. ## Spec sources read - `control-plane/platformctl/ci/pr_sanity.py`: model-review dispatch and deterministic findings. - `control-plane/platformctl/tests/test_pr_sanity.py`: existing PR sanity coverage and added regression tests. - `docs/forgejo-agent-operations.md`: Ollama Cloud review skip rules and Patchwarden boundary. - `decisions/0023-gemini-3-5-flash-reviewer-cousin.md`: Gemini/AntiGravity advisory lane. - `state/cycle/W6d-autonomous-merge-pilot.md`: Patchwarden model split and merge boundary. - Forgejo issues `#726`-`#731` via read-only API: wave target metadata and issue bodies. ## Verification ```text PYTHONPATH=control-plane uv run --project control-plane --extra dev python -m pytest control-plane/platformctl/tests/test_pr_sanity.py 35 passed python3 -m json.tool <wave json files> json ok rg secret-pattern scan over docs/agent-waves, state/agent-wave-template.md, and wave artifacts no matches git diff --check clean ``` ## Test plan - [ ] Confirm deterministic secret/trust-boundary blockers produce skipped model reviews. - [ ] Confirm safe docs diffs still dispatch configured model reviews. - [ ] Confirm wave JSON artifacts are valid and sanitized. - [ ] Run full security-sensitive review/canary before merge.
feat(agent-waves): pilot openai saver wave
All checks were successful
canary-required / collect-diff (pull_request) Successful in 5s
infra-docs-drift / docs-drift (pull_request) Successful in 5s
platformctl plan / auto-apply scope (pull_request) Successful in 19s
pyfallow / Pyfallow gate (control-plane) (pull_request) Successful in 18s
python-ci / Python 3.11 (pull_request) Successful in 45s
python-ci / Python 3.12 (pull_request) Successful in 45s
patchwarden-pr-sanity / collect-diff (pull_request) Successful in 4s
patchwarden-client-dry-run / dry-run (pull_request) Successful in 19s
python-ci / Python 3.13 (pull_request) Successful in 45s
workflow-lint / lint (pull_request) Successful in 5s
canary-required / canary (pull_request) Successful in 13s
base-is-main / guard (pull_request) Successful in 1s
patchwarden-client-dry-run / collect-diff (pull_request) Successful in 4s
patchwarden-pr-sanity / sanity (pull_request) Successful in 3m26s
f78aad1062
First-time contributor

Patchwarden PR sanity

  • Status: advisory_findings
  • PR: 746
  • Commit: f78aad1062453be90f4e894e0c849ac01e48bf8e
  • Security-sensitive label: present
  • Authority: advisory model review plus deterministic blockers only
  • 3+3 canary: still alive; this does not replace it

Deterministic findings

No deterministic findings.

Model reviewers

global-glm / glm-5.1:cloud

  • Status: ok

  • Verdict: OK

  • medium CLOUD_DISPATCH_BLOCKER_RULES may be incomplete for future blocker-severity findings

    • Evidence: control-plane/platformctl/ci/pr_sanity.py lines 70-80: CLOUD_DISPATCH_BLOCKER_RULES is an explicit frozenset of 8 ruleIds. If new blocker-severity findings are added to deterministic_findings() that could leak sensitive content to cloud mod
    • Next: Consider adding a code comment or docstring near CLOUD_DISPATCH_BLOCKER_RULES noting that new blocker-severity findings affecting secret/trust-boundary safety must be added here, or evaluate whether all blocker-severity findings should block cloud dispatch by default.

global-deepseek / deepseek-v4-pro:cloud

  • Status: ok

  • Verdict: OK

  • high Missing canary for security-sensitive PR

    • Evidence: PR description states 'Canary status: missing — fire canary 3+3 manually before merge'. The PR is labeled class/security-sensitive and the requested decision is 'approve_merge after full review/canary'. Merging without the canary would viol
    • Next: Complete the canary (3+3) before merging. Confirm canary results are clean and documented in the PR.

redteam / kimi-k2.6:cloud

  • Status: ok

  • Verdict: NOT_OK

  • blocker Workflow secret reference bypass via whitespace omission

    • Evidence: control-plane/platformctl/ci/pr_sanity.py FORBIDDEN_WORKFLOW_SNIPPETS tuples (e.g., "${{ secrets.FORGEJO_TOKEN") require literal whitespace after {{, but Forgejo/GitHub Actions expression syntax permits ${{secrets.FORGEJO_TOKEN}} withou
    • Next: Add whitespace-agnostic variants of the workflow patterns (e.g., "${{secrets.FORGEJO_TOKEN") or normalize diff content by removing spaces around {{ and secrets. before matching.

Policy notes

  • GLM 5.1 + DeepSeek V4 Pro are the operator-required model mix for this bot.
  • Optional red-team model is enabled only when PLATFORMCTL_PR_SANITY_REDTEAM_MODEL is configured.
  • Auto-merge is not enabled here.
<!-- patchwarden-pr-sanity:pdurlej/platform:PR-746 --> # Patchwarden PR sanity - Status: `advisory_findings` - PR: `746` - Commit: `f78aad1062453be90f4e894e0c849ac01e48bf8e` - Security-sensitive label: `present` - Authority: advisory model review plus deterministic blockers only - 3+3 canary: still alive; this does not replace it ## Deterministic findings No deterministic findings. ## Model reviewers ### `global-glm` / `glm-5.1:cloud` - Status: `ok` - Verdict: `OK` - **`medium`** CLOUD_DISPATCH_BLOCKER_RULES may be incomplete for future blocker-severity findings - Evidence: `control-plane/platformctl/ci/pr_sanity.py lines 70-80: CLOUD_DISPATCH_BLOCKER_RULES is an explicit frozenset of 8 ruleIds. If new blocker-severity findings are added to deterministic_findings() that could leak sensitive content to cloud mod` - Next: Consider adding a code comment or docstring near CLOUD_DISPATCH_BLOCKER_RULES noting that new blocker-severity findings affecting secret/trust-boundary safety must be added here, or evaluate whether all blocker-severity findings should block cloud dispatch by default. ### `global-deepseek` / `deepseek-v4-pro:cloud` - Status: `ok` - Verdict: `OK` - **`high`** Missing canary for security-sensitive PR - Evidence: `PR description states 'Canary status: missing — fire canary 3+3 manually before merge'. The PR is labeled class/security-sensitive and the requested decision is 'approve_merge after full review/canary'. Merging without the canary would viol` - Next: Complete the canary (3+3) before merging. Confirm canary results are clean and documented in the PR. ### `redteam` / `kimi-k2.6:cloud` - Status: `ok` - Verdict: `NOT_OK` - **`blocker`** Workflow secret reference bypass via whitespace omission - Evidence: `control-plane/platformctl/ci/pr_sanity.py FORBIDDEN_WORKFLOW_SNIPPETS tuples (e.g., "${{ secrets.FORGEJO_TOKEN") require literal whitespace after `{{`, but Forgejo/GitHub Actions expression syntax permits `${{secrets.FORGEJO_TOKEN}}` withou` - Next: Add whitespace-agnostic variants of the workflow patterns (e.g., "${{secrets.FORGEJO_TOKEN") or normalize diff content by removing spaces around `{{` and `secrets.` before matching. ## Policy notes - GLM 5.1 + DeepSeek V4 Pro are the operator-required model mix for this bot. - Optional red-team model is enabled only when `PLATFORMCTL_PR_SANITY_REDTEAM_MODEL` is configured. - Auto-merge is not enabled here.
pdurlej deleted branch codex/openai-saver-pilot 2026-06-08 15:03:39 +02:00
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
2 participants
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!746
No description provided.