/accept-adapt-reject

/accept-adapt-reject (alias: /aar)

A slash command that wraps the accept-adapt-reject skill. Pass any incoming feedback (code review comments, PR feedback, user corrections, design critiques) as the command arguments. The skill produces a structured AAR triage: each discrete piece of feedback is bucketed as ACCEPT, ADAPT, or REJECT, with reasoning and concrete actions. Accepted and adapted items are then executed.

Usage

/accept-adapt-reject <feedback text or reference to a thread/PR/comments>
/aar <feedback text or reference>

Examples

/aar "The function is too long. Rename tmp to temporaryValue. We should switch to GraphQL."
/aar PR #1234 review comments
/aar "Stop using const for everything, prefer let"

What it does

1. Loads the accept-adapt-reject skill for the full AAR algorithm, output template, and priority rules.
2. Segments the feedback into atomic items (one bullet/sentence/paragraph = one item).
3. For each item, gathers context (original intent, reviewer goal, constraints, prior accepted feedback) and applies the rubric: factually correct? serves the stated goal? is the proposed solution the best? cost of being wrong?
4. Outputs structured verdicts — ACCEPT / ADAPT / REJECT with reasoning and concrete actions, then a Summary block with counts and net changes.
5. Executes the accepted and adapted items (skips rejected ones, with a short rationale).

When to invoke

• After receiving a code review or PR review comment thread.
• When the user asks to "process this feedback", "triage these comments", or "decide what to do with this review".
• When multiple conflicting opinions need reconciliation.
• When the user expresses uncertainty about whether to apply feedback.

When NOT to invoke

• Pure compliments ("looks good!") — no triage needed.
• Questions, not feedback ("why did you choose X?") — answer the question directly instead.
• Bug reports that need debugging first — use systematic-debugging to find the root cause, then come back to AAR if feedback is involved.

Full skill

For the complete algorithm, priority rules, anti-patterns, and worked examples, see the canonical skill at:

~/.hermes_prod/skills/accept-adapt-reject/SKILL.md

The skill is the source of truth. This command is a thin launcher.