UNPKG

paqad-ai

Version:

Spec-driven development framework — AI agents that think before they type

github.com/Eliyce/paqad-ai

Eliyce/paqad-ai

38 lines (25 loc) • 1.63 kB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
# Re-run rules verified: 2026-11-02 A re-run is a re-check, not a re-audit with moved goalposts. Its whole value is that it answers one question honestly (does the stored map still match the code) using the exact verification the `site-map` workflow runs. These are the rules that keep it honest. ## One engine, one map - There is no separate retest engine and no report replay. The re-run is `paqad-ai sitemap run` over the one stored map at `docs/site-map/`, and its lasting record is the trust and freshness the engine stamps back into that map. ## Match by stable id, never by label - Finding ids are content-addressed (`SM-<hash8>`), so a finding that persists across runs keeps its id and the baseline ratchet marks it `pre-existing`. A renamed surface is the same finding; a coincidentally-similar new one is not. ## What drift means - `anchors_broken > 0` in the stamped freshness is the drift verdict: code the map cites no longer resolves. That is a finding against the map, never a reason to quietly drop the claim. - A finding that disappears between runs was resolved by whatever change removed its evidence, say which change, when you can see it, rather than just counting it gone. ## What a re-run never does - It never invents a finding. The engine's output is the only source. - It never softens a finding. Absence of proof is drift, not a fix. - It never hand-edits the stamped freshness. Only the verb writes it. ## The verdict Safe to merge only when the run exits clean and no cited anchor is broken. Blocked checks are a gap in confidence, not a pass: surface them, do not bury them.