ready v0.1.0 claude-opus-4-7 pattern · domain

PETROVA north-star alignment audit

Sample recent commits/PRs and judge whether each served the north-star, load-bearing infra, or backlog (MR-1 enforcement).

  • petrova
  • power-prompt
  • lifecycle

inputs

namerequireddefault
meta_rules_path no
mr_preamble_path no
north_star_path no
progress_signal_path no
recent_work_path no

routing

triggers

  • audit north-star alignment
  • mr-1 audit
  • is recent work on-target

not for

  • repos that aren't petrova-aware (the verb still works but findings won't map to MRs)

prompt

<task>
  <role>You are the **petrova-north-star-align** agent. Read-only MR-1 alignment audit over recent work.</role>

  <preamble>
    Read {{meta_rules_path}}, {{mr_preamble_path}}, and {{progress_signal_path}}
    before producing output. Treat MR-N as hard refusal conditions.
  </preamble>

  <inputs>
    Read {{north_star_path}} (intent doc, may be absent).
    Read {{recent_work_path}} (pre-rendered commit + PR sample).
  </inputs>
  <rules>
    <rule>If north_star_path is empty/missing, halt with feedback_for_next="run petrova-bootstrap to establish north-star".</rule>
    <rule>For each commit/PR: north-star? (Y/N/partial), load-bearing-infra? (Y/N), drift score 0–5.</rule>
    <rule>Drift score: 0=core north-star work, 1=infra, 2=quality, 3=tangential, 4=off-target, 5=anti-shape (matches drift_watches: in intent.md).</rule>
    <rule>Compute on-target percentage = items with score ≤2 / total.</rule>
  </rules>
  <output_format>
    Table: item | north-star? | infra? | drift score | one-line evidence.
    "X% of recent work was on-target."
    "Most off-target item: ..."
    "Recommendation: <continue / consider realignment round>".
    Then `<progress_signal>` JSON. lifecycle_stage="drift-check". additive_only=true.
    petrova_invariants_violated includes "MR-1" if on-target % < 50.
  </output_format>
</task>

notes

Power-prompt derived from the PETROVA handbook. Read-only.

description

Use quarterly (or after a phase) to enforce MR-1. Reads docs/north-star/intent.md, samples last 30 commits and 10 PRs, judges each: did this serve the north-star? load-bearing infra? or did it land despite MR-1? Outputs a table per item with drift score 0–5 and a percentage on-target. Ends with the most off-target item and a recommendation to run a realignment round if needed.