Context
When agents receive peer-reviews on their Pull Requests (e.g., from Claude to Gemini, or vice versa), the original "warm" context of the agent's initial PR design is often "cold" (interruption amnesia). This currently leads to agents passively complying with (rubber-stamping) change requests that may actively degrade the PR's original architectural intent.
We currently store the "Review Response Protocol" inside .agents/skills/pull-request/references/pull-request-workflow.md.
However, to reduce cognitive load and firmly establish an "Anti-Passive Compliance" evaluation mechanism, we need to extract this into its own dedicated skill reference document.
This aligns with our existing trajectory of applying the Progressive Disclosure Pattern to agent skills (as seen in #9986, #9703, and #9701).
Scope
- Extract and Isolate: Extract Section 7 ("Review Response Protocol") from
pull-request-workflow.md into a new, dedicated file: .agents/skills/pull-request/references/review-response-protocol.md.
- The 1-Liner Trigger: Add a lightweight trigger instruction in
AGENTS.md informing the agent that when receiving a review with "Request Changes" on their own PR, they must consult this new reference file.
- The Triangular Evaluation: The new
review-response-protocol.md will mandate a "Triangular Evaluation":
- Retrieve: Read the
Origin Session ID from the PR body.
- Re-hydrate: Query Memory Core (using
query_raw_memories or get_session_memories) for the original architectural intent.
- Evaluate & Defend: Compare the Original Intent against the Reviewer's Request and Fresh Analysis. If the reviewer's request contradicts the established architecture or original valid intent, the agent must defend the PR and use appropriate rejection markers (e.g.,
[REJECTED_WITH_RATIONALE]).
Acceptance Criteria
A2A Context
Context
When agents receive peer-reviews on their Pull Requests (e.g., from Claude to Gemini, or vice versa), the original "warm" context of the agent's initial PR design is often "cold" (interruption amnesia). This currently leads to agents passively complying with (rubber-stamping) change requests that may actively degrade the PR's original architectural intent.
We currently store the "Review Response Protocol" inside
.agents/skills/pull-request/references/pull-request-workflow.md. However, to reduce cognitive load and firmly establish an "Anti-Passive Compliance" evaluation mechanism, we need to extract this into its own dedicated skill reference document.This aligns with our existing trajectory of applying the Progressive Disclosure Pattern to agent skills (as seen in #9986, #9703, and #9701).
Scope
pull-request-workflow.mdinto a new, dedicated file:.agents/skills/pull-request/references/review-response-protocol.md.AGENTS.mdinforming the agent that when receiving a review with "Request Changes" on their own PR, they must consult this new reference file.review-response-protocol.mdwill mandate a "Triangular Evaluation":Origin Session IDfrom the PR body.query_raw_memoriesorget_session_memories) for the original architectural intent.[REJECTED_WITH_RATIONALE]).Acceptance Criteria
.agents/skills/pull-request/references/review-response-protocol.mdis created containing the extracted and enhanced protocol..agents/skills/pull-request/references/pull-request-workflow.mdhas Section 7 removed.AGENTS.mdhas a 1-liner hook instructing agents to read the new protocol when their PR receives change requests.A2A Context