UNPKG

paqad-ai

Version:

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

github.com/Eliyce/paqad-ai

Eliyce/paqad-ai

58 lines (42 loc) • 2.6 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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
--- name: site-map-retest description: Re-checks the stored site map against the current code by running the same verification again, never inventing or softening a finding and never calling missing proof a fix. Use when the user asks whether the site map still matches the code or its prior gaps are fixed. Only for the stored site map. metadata: model_tier: reasoning cacheable: 'false' output_format: json paqad-meta: skill.meta.yaml --- ## What It Does Drives the same deterministic verification the `site-map` workflow runs and reads what changed: the engine re-resolves every cited `file:line`, re-derives the findings, and re-stamps the stored map's earned trust tiers and its map-vs-code freshness. There is no separate retest engine and no report replay: a re-run is the same run again, and drift shows up as stamped proof in the map itself. ## Use This When When the `site-map` skill runs its Retest mode for the `site-map-retest` workflow, and calls this skill as a helper: the user wants to know whether the map still matches the code or prior gaps got fixed, not a fresh map. ## Inputs - The stored map at `docs/site-map/` (the one source of truth). - Read `references/replay-rules.md` before narrating any change. ## Procedure The verification is the engine's; you drive it and narrate honestly. 1. Run `paqad-ai sitemap run` (or hit Run on the dashboard's Site map area); it verifies the stored map against the current code and stamps the earned trust and freshness back into it. 2. Read the drift: `anchors_broken > 0` in the stamped freshness means cited code no longer resolves. Finding ids are content-addressed (`SM-<hash8>`), so a persisting finding keeps its id and the baseline ratchet marks it `pre-existing`; a finding that disappears was resolved by whatever change removed its evidence. 3. Narrate the split in the paqad voice: Safe to merge only when the run exits clean and no cited anchor is broken. ## Output Contract - A JSON object `{ findings: N, anchors_total: N, anchors_resolved: N, anchors_broken: N, new_since_baseline: N, pre_existing: N }`. - The verdict is Safe to merge only when `findings` is 0 and `anchors_broken` is 0. ## Escalate / Stop Conditions - Never invent a finding in a re-run; the engine's output is the only source. - Never soften a finding: absence of proof is drift, not a fix, a surface whose cited evidence no longer resolves is a finding, never silently fine. - Never hand-edit the stamped freshness; only the verb writes it. ## Resources - `references/replay-rules.md` - `agents/openai.yaml`