@antoneeo/kb-agentic-skill 1.3.0 → 1.4.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +51 -0
- package/README.md +89 -88
- package/gemini-extension.json +1 -1
- package/package.json +1 -1
- package/scripts/init.js +28 -0
- package/skills/kb-agentic-skill/SKILL.md +4 -2
- package/skills/kb-agentic-skill/scripts/sdlc_core.py +199 -3
- package/skills/kb-agentic-skill/templates.md +86 -33
package/CHANGELOG.md
CHANGED
|
@@ -2,6 +2,57 @@
|
|
|
2
2
|
|
|
3
3
|
Every significant change to this skill is recorded here.
|
|
4
4
|
|
|
5
|
+
## [1.4.0] - 2026-08-03
|
|
6
|
+
|
|
7
|
+
F-028 — several people on one project. `templates.md` had claimed the workstream
|
|
8
|
+
registry was "Parallel-safe by construction" since F-019, and nothing had ever exercised
|
|
9
|
+
it. Two workstreams opened from one base conflict **twice in one file**: on the row
|
|
10
|
+
insert, and on the file-global `Date:` header. Row-level ownership cannot save a file
|
|
11
|
+
that has a file-level field.
|
|
12
|
+
|
|
13
|
+
### Changed
|
|
14
|
+
- **`audit/handoff.md` is now GENERATED** from one `HANDOFF_[unit].md` per open
|
|
15
|
+
workstream, whose frontmatter IS the row. Two writers on two workstreams touch two
|
|
16
|
+
different files, and the `Date:` header is derived — from the newest `updated:` VALUE
|
|
17
|
+
in the sources, never a filesystem timestamp (git does not preserve mtimes, and an
|
|
18
|
+
mtime-derived header would make the file regenerate differently in every fresh clone).
|
|
19
|
+
- **`HANDOFF_[unit].md` is now written for every OPEN workstream**, with or without
|
|
20
|
+
volatile state — no file, no row. It still carries the resume logistics, and deleting
|
|
21
|
+
it at closure *is* what removes the row. The DRY boundary against the ANALYSIS Diary
|
|
22
|
+
is restated in the template, because what used to keep narrative out of that file was
|
|
23
|
+
its rarity, and the rarity is gone.
|
|
24
|
+
- Project-wide notes move to their own source, `audit/project_notes.md`, so generating
|
|
25
|
+
the registry cannot destroy notes that belong to no workstream.
|
|
26
|
+
|
|
27
|
+
### Added
|
|
28
|
+
- **`validate` errors when the registry disagrees with its sources** — which is what
|
|
29
|
+
turns a merge resolved carelessly from permanent into loud. Resolution is mechanical:
|
|
30
|
+
re-run `index`. The generated view can still conflict; the authored truth does not.
|
|
31
|
+
- **`index` refuses to write while anything in the file is unaccounted for**, and names
|
|
32
|
+
it. Converting one row at a time is the state that loses the others, so conversion is
|
|
33
|
+
per project. A project with no sources is untouched and sees no new finding.
|
|
34
|
+
- **`init` writes a `.gitattributes` stanza** giving the append-only review log
|
|
35
|
+
`merge=union` — a **built-in** driver, unlike `merge=ours`, which silently does
|
|
36
|
+
nothing until every clone runs `git config` and leaves the file wrong even then.
|
|
37
|
+
Create-only; a user's own `.gitattributes` is never clobbered.
|
|
38
|
+
- `scripts/test_merge_safety.py` (shared, ×3): the experiment that found the defect,
|
|
39
|
+
kept as a regression test — two workstreams from one base, merged, with the assertion
|
|
40
|
+
that only the generated view may conflict and that regenerating loses no state. Plus
|
|
41
|
+
the mixed-state, fresh-mtime, ordering, cap and duplicate-id guards, each
|
|
42
|
+
mutation-verified.
|
|
43
|
+
- A warning when two files claim the same workstream: the collision this design does
|
|
44
|
+
**not** fix (two people opening the same work under two names) must not pass as two
|
|
45
|
+
ordinary rows.
|
|
46
|
+
|
|
47
|
+
The mechanism is files and a generator: **it works with no VCS at all.** The
|
|
48
|
+
`.gitattributes` stanza is defence in depth — without it the outcome is today's, one
|
|
49
|
+
conflict resolved by hand, never a lost row.
|
|
50
|
+
|
|
51
|
+
### Upgrading
|
|
52
|
+
Nothing happens until you convert: with no `HANDOFF_*.md` sources, `index` and
|
|
53
|
+
`validate` behave exactly as before. When you convert, convert the whole registry at
|
|
54
|
+
once — `index` will tell you what is still unaccounted for.
|
|
55
|
+
|
|
5
56
|
## [1.3.0] - 2026-08-03
|
|
6
57
|
|
|
7
58
|
Field defect: an agent handed a 200-page manual emits a few dozen claims and reports
|
package/README.md
CHANGED
|
@@ -1,88 +1,89 @@
|
|
|
1
|
-
# KB Agentic Skill for Claude Code, Gemini CLI, Google Antigravity & Codex
|
|
2
|
-
|
|
3
|
-
`kb-agentic` turns an AI agent into the keeper of a second brain built **on the documents you supply** — not on what the model remembers. It supports Claude Code, Codex, Gemini CLI, Google Antigravity 2.0, Cursor/Windsurf-style project instructions, and optional devPNT governance.
|
|
4
|
-
|
|
5
|
-
The sibling of [`@antoneeo/agentic-sdlc-skill`](https://www.npmjs.com/package/@antoneeo/agentic-sdlc-skill), transplanted from source code to knowledge: same process, one different fidelity discipline.
|
|
6
|
-
|
|
7
|
-
## What it does
|
|
8
|
-
|
|
9
|
-
Two axes are kept apart on purpose — **abstraction** (a topic made of topics: the graph edges) and **certainty** (how strongly the corpus supports a statement: the claim rows).
|
|
10
|
-
|
|
11
|
-
1. **Intake — the source becomes untouchable.** Every document enters verbatim under `corpus/given/`, content-addressed by a raw-byte digest. A new version is appended with `supersedes:`, never overwritten. Non-text files get a **stored canonical extraction** (`.txt`, registered extractor) — those are the bytes locators point at, so an id stays valid across re-ingestion. On a large binary corpus the extraction **is** the artifact: the digest moves onto it (immutability still enforced, on the bytes that actually matter) and the original stays where it lives, recorded as `original_path:` + `original_sha256:`. 233 MB of manuals do not enter your repository to protect bytes no locator addresses — and the two recorded fields are honest about their limit: they let a human re-verify, they check nothing on their own. Spoken input is a note with `origin: elicited`; a synthesis carries `derived_from:`; your decision carries `basis:`. A note with none of the three is model knowledge dressed as a source, and the validator rejects it.
|
|
12
|
-
|
|
13
|
-
2. **Extraction — the unit is the claim.** Rows of `id | claim | valid | qty | about | source | prov | state`. The **id hashes the location and the quantity, never the text**, so an LLM rephrasing mints no new identity. The locator (`p=17@412-509`) is verified: the validator opens the extraction and checks the span exists — and `anchor <path> <phrase>` produces it for you, matching spaces as `\s+` because a PDF extraction breaks phrases mid-line. **Gates are extracted alongside powers**: for every row saying what something can do, the source is asked what must hold first — default-off, licence tier, version floor, dependency — because "yes, supported" without the gate is a plan that fails on site. The rule is *ask*, never *produce*: a source that states no gate yields no row. Validity scopes are half-open ("until March" and "from March" do not conflict). Quantities are typed — mixed kinds or currencies **refuse to sum**. What the source does not assert becomes a `gaps:` line, never a claim.
|
|
14
|
-
|
|
15
|
-
**The source is exhausted, not sampled.** "Invents nothing" is a floor, and an extractor that stops when nothing it wrote is false stops on page twenty of a two-hundred-page manual with every row correct — which is exactly what a 200-page manual produces in the field. So a long source is read in **bounded windows** (30 pages by default, one plan task each, the ledger holding your place across sessions), and every window closes by advancing `extracted_through:` on the artifact's sidecar. That field is what makes "I am finished" falsifiable: claims with no coverage recorded are an error, a claim addressing a page past the declared coverage is a contradiction, and coverage short of the end is reported until it reaches it. Its limit is stated where it is written — **nothing proves a page was read**; what changes is that the shortcut must now be written down to pass. And exhaustive means read, never *a row per page*: a page that asserts nothing yields nothing.
|
|
16
|
-
|
|
17
|
-
3. **Placement — five verdicts, after querying the graph.** Descent through the generated index following every parent (polyhierarchy). EXISTS → reconcile; INADEQUATE → child; **MISSING only after the graph was actually asked**; GENERALIZES → escalate (a new root stops at you); UNPLACED → quarantine. Similar-but-maybe-different becomes a sibling **with the distinguishing line written** — if you cannot write it, it is the same concept. Cycles are refused at write time; merged nodes leave a tombstone with `redirect_to:`, never a deletion.
|
|
18
|
-
|
|
19
|
-
4. **Reconciliation — the machine detects and holds, it never decides.** Five outcomes: new / confirmation (the source is appended to the row — the base strengthens, it does not lengthen) / refinement (the old row goes `SUPERSEDED`, its text intact) / coexistence (disjoint scopes) / conflict → the whole set goes `CONTESTED`, **symmetrically**: flipping one cell by hand fails the check. Only new information resolves it — a later source, or **your ruling with a `basis:`**, the fact you know and the corpus does not. No basis, no ruling: a preference is not a fact. A ruling is challengeable — a later document reopens the case with your basis beside it.
|
|
20
|
-
|
|
21
|
-
5. **Escalation in one batch at the end of the run**, in legal form (the claims, the reopenable sources, the dates, why the machine cannot decide). Ingestion never stops to interrogate you.
|
|
22
|
-
|
|
23
|
-
Deliberately absent: any per-node coverage or completion state. `gaps:` says what a node lacks; nothing collects it into a dashboard. A **source** does record how far it has been read, but on its own sidecar and nowhere else — the corpus index prints that fact for every artifact, including the finished ones, because a list of only what is behind is the dashboard this method refuses.
|
|
24
|
-
|
|
25
|
-
## Key features
|
|
26
|
-
|
|
27
|
-
- **Risk-proportional triage, measured in knowledge and never in files**: one claim row → propagating a fact already settled → a new knowledge unit (a source ingested, a node created or superseded, the hierarchy moved). Carrying one settled fact into eight documents is small; one claim that re-parents a node is not. A **Write Triggers** table maps each knowledge event to exactly one destination.
|
|
28
|
-
- **Portable knowledge**: `export` bundles a subgraph together with the bytes its claims cite — a closure, not a selection, because a claim whose source cannot be reopened is model knowledge arriving by another route. `import` merges it into another project **additively**: it never overwrites a node and never deletes, and claims already present are recognised by id rather than by comparing text, since the same artifact cited at the same span mints the same id in every project. Knowledge crosses the project boundary; **authority does not** — an imported ruling arrives as `prov: IMPORTED`, keeps its original `basis:`, and cannot settle a local disagreement until you re-ratify it.
|
|
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
|
-
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
62
|
-
|
|
63
|
-
|
|
64
|
-
|
|
65
|
-
|
|
66
|
-
|
|
67
|
-
|
|
68
|
-
|
|
69
|
-
|
|
70
|
-
|
|
71
|
-
|
|
72
|
-
|
|
73
|
-
|
|
74
|
-
| `@antoneeo/
|
|
75
|
-
|
|
|
76
|
-
|
|
77
|
-
|
|
78
|
-
|
|
79
|
-
|
|
80
|
-
|
|
81
|
-
|
|
82
|
-
- **
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
86
|
-
|
|
87
|
-
|
|
88
|
-
|
|
1
|
+
# KB Agentic Skill for Claude Code, Gemini CLI, Google Antigravity & Codex
|
|
2
|
+
|
|
3
|
+
`kb-agentic` turns an AI agent into the keeper of a second brain built **on the documents you supply** — not on what the model remembers. It supports Claude Code, Codex, Gemini CLI, Google Antigravity 2.0, Cursor/Windsurf-style project instructions, and optional devPNT governance.
|
|
4
|
+
|
|
5
|
+
The sibling of [`@antoneeo/agentic-sdlc-skill`](https://www.npmjs.com/package/@antoneeo/agentic-sdlc-skill), transplanted from source code to knowledge: same process, one different fidelity discipline.
|
|
6
|
+
|
|
7
|
+
## What it does
|
|
8
|
+
|
|
9
|
+
Two axes are kept apart on purpose — **abstraction** (a topic made of topics: the graph edges) and **certainty** (how strongly the corpus supports a statement: the claim rows).
|
|
10
|
+
|
|
11
|
+
1. **Intake — the source becomes untouchable.** Every document enters verbatim under `corpus/given/`, content-addressed by a raw-byte digest. A new version is appended with `supersedes:`, never overwritten. Non-text files get a **stored canonical extraction** (`.txt`, registered extractor) — those are the bytes locators point at, so an id stays valid across re-ingestion. On a large binary corpus the extraction **is** the artifact: the digest moves onto it (immutability still enforced, on the bytes that actually matter) and the original stays where it lives, recorded as `original_path:` + `original_sha256:`. 233 MB of manuals do not enter your repository to protect bytes no locator addresses — and the two recorded fields are honest about their limit: they let a human re-verify, they check nothing on their own. Spoken input is a note with `origin: elicited`; a synthesis carries `derived_from:`; your decision carries `basis:`. A note with none of the three is model knowledge dressed as a source, and the validator rejects it.
|
|
12
|
+
|
|
13
|
+
2. **Extraction — the unit is the claim.** Rows of `id | claim | valid | qty | about | source | prov | state`. The **id hashes the location and the quantity, never the text**, so an LLM rephrasing mints no new identity. The locator (`p=17@412-509`) is verified: the validator opens the extraction and checks the span exists — and `anchor <path> <phrase>` produces it for you, matching spaces as `\s+` because a PDF extraction breaks phrases mid-line. **Gates are extracted alongside powers**: for every row saying what something can do, the source is asked what must hold first — default-off, licence tier, version floor, dependency — because "yes, supported" without the gate is a plan that fails on site. The rule is *ask*, never *produce*: a source that states no gate yields no row. Validity scopes are half-open ("until March" and "from March" do not conflict). Quantities are typed — mixed kinds or currencies **refuse to sum**. What the source does not assert becomes a `gaps:` line, never a claim.
|
|
14
|
+
|
|
15
|
+
**The source is exhausted, not sampled.** "Invents nothing" is a floor, and an extractor that stops when nothing it wrote is false stops on page twenty of a two-hundred-page manual with every row correct — which is exactly what a 200-page manual produces in the field. So a long source is read in **bounded windows** (30 pages by default, one plan task each, the ledger holding your place across sessions), and every window closes by advancing `extracted_through:` on the artifact's sidecar. That field is what makes "I am finished" falsifiable: claims with no coverage recorded are an error, a claim addressing a page past the declared coverage is a contradiction, and coverage short of the end is reported until it reaches it. Its limit is stated where it is written — **nothing proves a page was read**; what changes is that the shortcut must now be written down to pass. And exhaustive means read, never *a row per page*: a page that asserts nothing yields nothing.
|
|
16
|
+
|
|
17
|
+
3. **Placement — five verdicts, after querying the graph.** Descent through the generated index following every parent (polyhierarchy). EXISTS → reconcile; INADEQUATE → child; **MISSING only after the graph was actually asked**; GENERALIZES → escalate (a new root stops at you); UNPLACED → quarantine. Similar-but-maybe-different becomes a sibling **with the distinguishing line written** — if you cannot write it, it is the same concept. Cycles are refused at write time; merged nodes leave a tombstone with `redirect_to:`, never a deletion.
|
|
18
|
+
|
|
19
|
+
4. **Reconciliation — the machine detects and holds, it never decides.** Five outcomes: new / confirmation (the source is appended to the row — the base strengthens, it does not lengthen) / refinement (the old row goes `SUPERSEDED`, its text intact) / coexistence (disjoint scopes) / conflict → the whole set goes `CONTESTED`, **symmetrically**: flipping one cell by hand fails the check. Only new information resolves it — a later source, or **your ruling with a `basis:`**, the fact you know and the corpus does not. No basis, no ruling: a preference is not a fact. A ruling is challengeable — a later document reopens the case with your basis beside it.
|
|
20
|
+
|
|
21
|
+
5. **Escalation in one batch at the end of the run**, in legal form (the claims, the reopenable sources, the dates, why the machine cannot decide). Ingestion never stops to interrogate you.
|
|
22
|
+
|
|
23
|
+
Deliberately absent: any per-node coverage or completion state. `gaps:` says what a node lacks; nothing collects it into a dashboard. A **source** does record how far it has been read, but on its own sidecar and nowhere else — the corpus index prints that fact for every artifact, including the finished ones, because a list of only what is behind is the dashboard this method refuses.
|
|
24
|
+
|
|
25
|
+
## Key features
|
|
26
|
+
|
|
27
|
+
- **Risk-proportional triage, measured in knowledge and never in files**: one claim row → propagating a fact already settled → a new knowledge unit (a source ingested, a node created or superseded, the hierarchy moved). Carrying one settled fact into eight documents is small; one claim that re-parents a node is not. A **Write Triggers** table maps each knowledge event to exactly one destination.
|
|
28
|
+
- **Portable knowledge**: `export` bundles a subgraph together with the bytes its claims cite — a closure, not a selection, because a claim whose source cannot be reopened is model knowledge arriving by another route. `import` merges it into another project **additively**: it never overwrites a node and never deletes, and claims already present are recognised by id rather than by comparing text, since the same artifact cited at the same span mints the same id in every project. Knowledge crosses the project boundary; **authority does not** — an imported ruling arrives as `prov: IMPORTED`, keeps its original `basis:`, and cannot settle a local disagreement until you re-ratify it.
|
|
29
|
+
- **Several people, one project**: the workstream registry (`audit/handoff.md`) is **generated** from one file per open workstream, so two people opening or closing two workstreams on two branches edit two different files and their merge is clean. Row-per-workstream alone was not enough — a file-global `Date:` header defeats row-level ownership — so the header is derived from the sources and no writer touches it. The generated view can still conflict; that conflict is resolved by re-running `index`, never by hand, and `validate` refuses CLEAN until the file matches its sources. The append-only review log gets `merge=union` (a built-in driver, no per-clone configuration). It all works with no VCS at all: it is files and a generator.
|
|
30
|
+
- **Vision-guided governance**: Standalone projects use `ai_docs/vision/`; Hybrid projects use devPNT `M-VISION` as the milestone north star. `DRAFT` informs, `APPROVED` binds, promotion is the user's alone.
|
|
31
|
+
- **Independent review, twice**: the design before it is implemented, the result before it is declared done — fresh-context subagent > one-shot run > a declared self-pass, 3 rounds max, one log line each, and a PASS is invalid on "found nothing".
|
|
32
|
+
- **Question discipline**: a question is legal only when the agent searched first, names the search with its result, and names the decision it unblocks; otherwise it proceeds on a declared assumption, batched.
|
|
33
|
+
- **Operative guides + agent-global KB**: distil user-provided indications into source-faithful `GUIDE_*.md` (`source_kind: document`) — verbatim snapshot plus hash, so drift is detected mechanically.
|
|
34
|
+
- **Mechanical checks**: `check`, `validate`, `index`, `graph`, `corpus`, `claim-id`, `anchor`, `export`, `import`, plus the spine's `stale`/`mark`/`gate`/`plan`/`orient`/`migrate`. The graph and corpus checks verify spans against the stored extraction, recompute every id, refuse cycles and unreachable nodes, and enforce `CONTESTED` symmetry.
|
|
35
|
+
- **Installed support files**: Claude, Codex, Gemini and Google Antigravity receive the full skill folder — `SKILL.md`, `templates.md`, `taxonomy.md`, `distillation.md`, `reconciliation.md`, `guides.md`, `vision.md`, `elicitation.md`, `review.md`, `dispatch.md`, `routing.md`, `ENFORCEMENT.md`, and the validator's two files, `scripts/sdlc_check.py` + `scripts/sdlc_core.py` (the core is the family's shared spine — copy both, or neither).
|
|
36
|
+
|
|
37
|
+
## Installation
|
|
38
|
+
|
|
39
|
+
```bash
|
|
40
|
+
npm install -g @antoneeo/kb-agentic-skill@latest
|
|
41
|
+
```
|
|
42
|
+
|
|
43
|
+
That is enough — the package's `postinstall` runs the installer. If your npm blocks
|
|
44
|
+
install scripts (`--ignore-scripts`, some CI/pnpm setups), run it by hand:
|
|
45
|
+
|
|
46
|
+
```bash
|
|
47
|
+
kb-agentic-install-skill
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
> The command is on your PATH only after a **global** (`-g`) install; after a local
|
|
51
|
+
> `npm i`, invoke it as `npx kb-agentic-install-skill`.
|
|
52
|
+
|
|
53
|
+
The installer copies `skills/kb-agentic-skill/` recursively into native skill locations:
|
|
54
|
+
|
|
55
|
+
- Claude Code: `~/.claude/skills/kb-agentic/`
|
|
56
|
+
- Codex: `~/.codex/skills/kb-agentic/`
|
|
57
|
+
- Gemini CLI: `~/.gemini/skills/kb-agentic/`
|
|
58
|
+
- Google Antigravity: `~/.gemini/config/skills/kb-agentic/` (override the home with `ANTIGRAVITY_HOME`)
|
|
59
|
+
|
|
60
|
+
Restart the relevant agent, or reload skills where the CLI supports it.
|
|
61
|
+
|
|
62
|
+
Initialize a project:
|
|
63
|
+
|
|
64
|
+
```bash
|
|
65
|
+
kb-agentic-init
|
|
66
|
+
```
|
|
67
|
+
|
|
68
|
+
Run it inside a project to create `ai_docs/`, Vision documents, strategic docs, audit plan, and agent protocol files (`AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, `.cursorrules`).
|
|
69
|
+
|
|
70
|
+
## The family: three lenses, one spine
|
|
71
|
+
|
|
72
|
+
| Package | Faithful to | Unit of work |
|
|
73
|
+
|---|---|---|
|
|
74
|
+
| [`@antoneeo/agentic-sdlc-skill`](https://www.npmjs.com/package/@antoneeo/agentic-sdlc-skill) | this repository's code | feature |
|
|
75
|
+
| `@antoneeo/kb-agentic-skill` (this one) | the documents you supply | topic |
|
|
76
|
+
| [`@antoneeo/mkt-agentic-sdlc-skill`](https://www.npmjs.com/package/@antoneeo/mkt-agentic-sdlc-skill) | market evidence | engagement |
|
|
77
|
+
|
|
78
|
+
Triage, the Vision Gate, the review gates, the guide router, question discipline and the validator spine are byte-identical across the three. When two live in the same project, `routing.md` decides which lens owns a given piece of work, and any of the three validators gives the same verdict on the same tree.
|
|
79
|
+
|
|
80
|
+
## Standalone vs Hybrid
|
|
81
|
+
|
|
82
|
+
- **Standalone** — `ai_docs/` is the source of truth: vision, topics, corpus, audit, handoff.
|
|
83
|
+
- **Hybrid with devPNT** — devPNT governs `M-VISION`, Master Plan, Action Plan and versioned artifacts; `ai_docs/` stays as readable context, fallback and shadow. Divergence between your request, the local Vision and the M-VISION is surfaced before any work.
|
|
84
|
+
|
|
85
|
+
## Created By
|
|
86
|
+
|
|
87
|
+
Created by **Antonio Pinto** ([GitHub](https://github.com/Antoneeo)).
|
|
88
|
+
|
|
89
|
+
MIT (c) 2026 Antonio Pinto.
|
package/gemini-extension.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "kb-agentic-skill",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.4.0",
|
|
4
4
|
"description": "Knowledge-Base & Document-First protocol with risk triage, Vision governance, signal distillation and optional devPNT integration.",
|
|
5
5
|
"author": "Antonio Pinto (https://github.com/Antoneeo)"
|
|
6
6
|
}
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@antoneeo/kb-agentic-skill",
|
|
3
|
-
"version": "1.
|
|
3
|
+
"version": "1.4.0",
|
|
4
4
|
"description": "Knowledge-Base & Document-First protocol for Claude Code, Gemini CLI, Google Antigravity and Codex with risk triage, Vision governance, signal distillation and optional devPNT integration.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"claude-code",
|
package/scripts/init.js
CHANGED
|
@@ -105,6 +105,34 @@ const writeIfNotExists = (relPath, content, description) => {
|
|
|
105
105
|
|
|
106
106
|
seedFiles.forEach(([relPath, content]) => writeIfNotExists(relPath, content));
|
|
107
107
|
|
|
108
|
+
const EOL = String.fromCharCode(10);
|
|
109
|
+
// 5b. Merge hygiene for the append-only review log (F-028, defence in depth).
|
|
110
|
+
// `merge=union` is a BUILT-IN driver: no per-clone `git config`, unlike
|
|
111
|
+
// `merge=ours`, which silently does nothing until every clone configures it.
|
|
112
|
+
// Rows are date-stamped, so interleaving them loses no information and their
|
|
113
|
+
// order carries none. Create-only, and appended rather than rewritten: a user's
|
|
114
|
+
// own .gitattributes is never clobbered.
|
|
115
|
+
const gitattributes = () => {
|
|
116
|
+
const marker = 'ai_docs/audit/reviews/REVIEW_LOG.md';
|
|
117
|
+
const stanza = [
|
|
118
|
+
'',
|
|
119
|
+
'# agentic-sdlc: the review log is append-only and its rows are date-stamped,',
|
|
120
|
+
'# so a union merge keeps both sides instead of asking a human to pick one.',
|
|
121
|
+
marker + ' merge=union',
|
|
122
|
+
'',
|
|
123
|
+
].join(EOL);
|
|
124
|
+
const filePath = path.join(cwd, '.gitattributes');
|
|
125
|
+
if (!fs.existsSync(path.join(cwd, '.git'))) return; // not a git repo: nothing to do
|
|
126
|
+
const current = fs.existsSync(filePath) ? fs.readFileSync(filePath, 'utf8') : '';
|
|
127
|
+
if (current.includes(marker)) {
|
|
128
|
+
console.log('SKIP .gitattributes already covers the review log.');
|
|
129
|
+
return;
|
|
130
|
+
}
|
|
131
|
+
fs.writeFileSync(filePath, current + stanza, 'utf8');
|
|
132
|
+
console.log('OK .gitattributes: review log set to merge=union (concurrent reviews merge cleanly).');
|
|
133
|
+
};
|
|
134
|
+
gitattributes();
|
|
135
|
+
|
|
108
136
|
// 6. Client discovery and protocol pointers
|
|
109
137
|
console.log('\n--- Environment Analysis ---');
|
|
110
138
|
|
|
@@ -74,8 +74,10 @@ Triage decides IF documentation is due; this table decides WHICH document each e
|
|
|
74
74
|
| `solutions/ANALYSIS_[topic].md` | Every L3, after elicitation and before any drafting. On topic match with an existing analysis, update that one instead of a new file. | 3 |
|
|
75
75
|
| `solutions/SPIKE_[topic].md` | Closing any Spike — including a negative outcome. | — |
|
|
76
76
|
| `vision/features/VISION_[feature].md` | Multi-milestone topic at analysis time, or retroactive trigger (second ANALYSIS on same theme). | 3 |
|
|
77
|
-
| `audit/handoff.md` (workstream registry) |
|
|
78
|
-
| `audit/HANDOFF_[topic].md` |
|
|
77
|
+
| `audit/handoff.md` (workstream registry) | **Never by hand — generated by `sdlc_check.py index` from the `HANDOFF_*.md` sources**, and `validate` errors when the two disagree. Regenerate at every L3 closure and at session end; the `Date:` header is derived, so no writer touches it. Inventory for lookup, never a work board. | 5 / session end |
|
|
78
|
+
| `audit/HANDOFF_[topic].md` | **One per OPEN workstream, with or without volatile state** — it is the authored home of that workstream's registry row (frontmatter `workstream`/`level`/`branch`/`status`/`since`/`next`/`details`/`updated`), so no file means no row. Carries the resume logistics too; the ANALYSIS Diary keeps the durable narrative (DRY). DELETED at closure — deleting it *is* removing the row. | 4 / 5 / session end |
|
|
79
|
+
| `audit/project_notes.md` | A note true for the whole project rather than one workstream. Appended verbatim to the generated registry, so regenerating cannot destroy notes that belong to no workstream. | 5 / session end |
|
|
80
|
+
| `audit/handoff.md` — converting an existing project | **Lazily, at the first write — and then ALL AT ONCE.** Converting one row at a time is the state that loses the others, so `index` refuses to write while anything in the file is unaccounted for and names it. Until the first source exists nothing generates and nothing errors. | 5 / session end |
|
|
79
81
|
| `audit/audit_plan.md` (Standalone) | Bootstrap, and whenever a mapped area changes state (`sdlc_check.py mark`). | 1 |
|
|
80
82
|
| `reference/GUIDE_[topic].md` (`source_kind: document`) | Origin+purpose test (`guides.md`), or a proactive proposal accepted by user. | 4 / 5 |
|
|
81
83
|
| `reference/GUIDE_[topic].md` (`source_kind: code`/`domain`) | Recognized high-complexity domain/concept with no CURRENT guide $\rightarrow$ duty to write autonomously. | 4 / 5 |
|
|
@@ -215,7 +215,11 @@ DEFAULT_DOMAIN = "code"
|
|
|
215
215
|
PORTABLE_CHECKS = {}
|
|
216
216
|
# Which check namespaces this distribution actually carries. Set by the entry point;
|
|
217
217
|
# a `checks:` entry outside it WARNS visibly rather than passing silently.
|
|
218
|
-
_ENTRY_POINT = {"domain": DEFAULT_DOMAIN, "provides": ()}
|
|
218
|
+
_ENTRY_POINT = {"domain": DEFAULT_DOMAIN, "provides": (), "script": "sdlc_check.py"}
|
|
219
|
+
# "script" is what generated headers tell the reader to RUN. Declared, never
|
|
220
|
+
# derived from sys.argv: the generated bytes must not depend on how the command
|
|
221
|
+
# was invoked, or the alignment check would fail on the invocation instead of on
|
|
222
|
+
# the content.
|
|
219
223
|
|
|
220
224
|
|
|
221
225
|
# --- distribution profile ----------------------------------------------------
|
|
@@ -292,10 +296,17 @@ def portable_check(name):
|
|
|
292
296
|
return register
|
|
293
297
|
|
|
294
298
|
|
|
295
|
-
def set_entry_point(domain, provides=()):
|
|
299
|
+
def set_entry_point(domain, provides=(), script=None):
|
|
296
300
|
"""Declare which domain this distribution is and which check namespaces it ships."""
|
|
297
301
|
_ENTRY_POINT["domain"] = domain
|
|
298
302
|
_ENTRY_POINT["provides"] = tuple(provides)
|
|
303
|
+
if script:
|
|
304
|
+
_ENTRY_POINT["script"] = script
|
|
305
|
+
|
|
306
|
+
|
|
307
|
+
def entry_script():
|
|
308
|
+
"""The command name a generated header tells the reader to run."""
|
|
309
|
+
return _ENTRY_POINT["script"]
|
|
299
310
|
|
|
300
311
|
|
|
301
312
|
def project_default_domain(root):
|
|
@@ -727,6 +738,142 @@ def build_manifest(root):
|
|
|
727
738
|
return "\n".join(lines).rstrip() + "\n"
|
|
728
739
|
|
|
729
740
|
|
|
741
|
+
# ------------------------------------------------- workstream registry (F-028)
|
|
742
|
+
# audit/handoff.md is GENERATED from one source file per open workstream, so two
|
|
743
|
+
# writers working two workstreams touch two different files. Row-per-workstream
|
|
744
|
+
# alone was not enough -- F-019 had that and the file still conflicted twice,
|
|
745
|
+
# because a file-global `Date:` header defeats row-level ownership. The header is
|
|
746
|
+
# now DERIVED, so no writer touches it.
|
|
747
|
+
|
|
748
|
+
REGISTRY_COLUMNS = ("Workstream", "Level", "Branch", "Status", "Since",
|
|
749
|
+
"Next step", "Details")
|
|
750
|
+
REGISTRY_KEYS = ("workstream", "level", "branch", "status", "since", "next")
|
|
751
|
+
REGISTRY_CAP = 20 # a signal, never a truncation (see cmd_index)
|
|
752
|
+
PROJECT_NOTES = "project_notes.md" # NOT handoff_notes.md: the HANDOFF_*.md
|
|
753
|
+
# glob is case-insensitive on Windows, and that name would be collected as a
|
|
754
|
+
# source. The trap is real; the name is the fix.
|
|
755
|
+
|
|
756
|
+
|
|
757
|
+
def registry_header():
|
|
758
|
+
return (f"<!-- GENERATED by {entry_script()} index - do not edit by hand. "
|
|
759
|
+
f"Source of truth: the HANDOFF_*.md files in {docs_dir()}/audit/. -->")
|
|
760
|
+
|
|
761
|
+
|
|
762
|
+
def list_workstreams(root):
|
|
763
|
+
"""[(path, meta)] for audit/HANDOFF_*.md carrying registry frontmatter.
|
|
764
|
+
|
|
765
|
+
Opt-in by the presence of `workstream:` -- a project whose handoff is still
|
|
766
|
+
hand-written has no sources, so nothing generates and nothing errors (the
|
|
767
|
+
F-019 migration lesson). Sorted by workstream id: the alignment check is a
|
|
768
|
+
byte comparison, and glob order differs across filesystems."""
|
|
769
|
+
aud = ai_path(root, "audit")
|
|
770
|
+
out = []
|
|
771
|
+
if not aud.is_dir():
|
|
772
|
+
return out
|
|
773
|
+
for p in sorted(aud.glob("HANDOFF_*.md")):
|
|
774
|
+
meta = load_frontmatter(read_text(p).splitlines())
|
|
775
|
+
if str(meta.get("workstream") or "").strip():
|
|
776
|
+
out.append((p, meta))
|
|
777
|
+
out.sort(key=lambda pm: (str(pm[1].get("workstream")).strip().lower(), pm[0].name))
|
|
778
|
+
return out
|
|
779
|
+
|
|
780
|
+
|
|
781
|
+
def _registry_cell(value):
|
|
782
|
+
text = str(value if value is not None else "").strip()
|
|
783
|
+
return text.replace("|", "\\|") or "-"
|
|
784
|
+
|
|
785
|
+
|
|
786
|
+
def build_registry(root):
|
|
787
|
+
"""The generated registry, or "" when there is no source to build it from.
|
|
788
|
+
|
|
789
|
+
Deterministic by construction: sorted rows, and a `Date:` taken from the
|
|
790
|
+
newest `updated:` VALUE written inside the sources -- never a filesystem
|
|
791
|
+
timestamp. Git does not preserve mtimes, so an mtime-derived header would
|
|
792
|
+
regenerate differently in every fresh clone and the alignment check would
|
|
793
|
+
fire on a tree nobody touched."""
|
|
794
|
+
rows = list_workstreams(root)
|
|
795
|
+
if not rows:
|
|
796
|
+
return ""
|
|
797
|
+
stamp = max(str(m.get("updated") or m.get("since") or "").strip() or "0000-00-00"
|
|
798
|
+
for _p, m in rows)
|
|
799
|
+
lines = ["# Handoff — workstream registry",
|
|
800
|
+
f"Date: {stamp} (UTC)",
|
|
801
|
+
"",
|
|
802
|
+
registry_header(),
|
|
803
|
+
"",
|
|
804
|
+
"| " + " | ".join(REGISTRY_COLUMNS) + " |",
|
|
805
|
+
"|" + "---|" * len(REGISTRY_COLUMNS)]
|
|
806
|
+
for p, meta in rows:
|
|
807
|
+
extra = str(meta.get("details") or "").strip()
|
|
808
|
+
details = f"{p.name} · {extra}" if extra else p.name
|
|
809
|
+
cells = [_registry_cell(meta.get(k)) for k in REGISTRY_KEYS]
|
|
810
|
+
lines.append("| " + " | ".join(cells) + f" | {_registry_cell(details)} |")
|
|
811
|
+
notes = ai_path(root, "audit", PROJECT_NOTES)
|
|
812
|
+
if notes.is_file():
|
|
813
|
+
body = read_text(notes).strip()
|
|
814
|
+
if body:
|
|
815
|
+
lines += ["", "## Project-wide notes", "", body]
|
|
816
|
+
return "\n".join(lines) + "\n"
|
|
817
|
+
|
|
818
|
+
|
|
819
|
+
def parse_registry_rows(text):
|
|
820
|
+
"""Workstream ids in a registry table, hand-written or generated."""
|
|
821
|
+
ids = []
|
|
822
|
+
for line in text.splitlines():
|
|
823
|
+
s = line.strip()
|
|
824
|
+
if not s.startswith("|") or set(s) <= {"|", "-", " ", ":"}:
|
|
825
|
+
continue
|
|
826
|
+
first = s.strip("|").split("|")[0].strip()
|
|
827
|
+
if first and first.lower() != "workstream":
|
|
828
|
+
ids.append(first)
|
|
829
|
+
return ids
|
|
830
|
+
|
|
831
|
+
|
|
832
|
+
def registry_conversion_blockers(root):
|
|
833
|
+
"""What stops `index` from writing over a hand-written handoff.md.
|
|
834
|
+
|
|
835
|
+
The mixed state is the trap this exists for: converting one row at a time
|
|
836
|
+
leaves a project at one source file and five hand-written rows, and
|
|
837
|
+
regenerating from the one source DELETES the other five -- silently, in the
|
|
838
|
+
file whose whole purpose is not losing them. So conversion is per project.
|
|
839
|
+
Empty list = writing is safe."""
|
|
840
|
+
hand = ai_path(root, "audit", "handoff.md")
|
|
841
|
+
if not hand.is_file():
|
|
842
|
+
return []
|
|
843
|
+
text = read_text(hand)
|
|
844
|
+
if "GENERATED by sdlc_check.py index" in text:
|
|
845
|
+
return [] # already ours
|
|
846
|
+
blockers = []
|
|
847
|
+
known = {str(m.get("workstream")).strip() for _p, m in list_workstreams(root)}
|
|
848
|
+
orphans = [r for r in parse_registry_rows(text) if r not in known]
|
|
849
|
+
if orphans:
|
|
850
|
+
blockers.append("rows no HANDOFF_*.md accounts for: " + ", ".join(orphans))
|
|
851
|
+
# Everything else in the file must have a home too, or it is lost on write:
|
|
852
|
+
# a pre-1.17 narrative handoff carries no table at all, so orphan rows alone
|
|
853
|
+
# would not notice it.
|
|
854
|
+
notes_ok = ai_path(root, "audit", PROJECT_NOTES).is_file()
|
|
855
|
+
leftovers, in_notes = [], False
|
|
856
|
+
for line in text.splitlines():
|
|
857
|
+
s = line.strip()
|
|
858
|
+
if not s or s.startswith("|") or s.startswith("<!--"):
|
|
859
|
+
continue
|
|
860
|
+
if s.startswith("# ") or re.match(r"^(?:Date|Data):", s):
|
|
861
|
+
continue
|
|
862
|
+
if re.match(r"^##\s+Project-wide notes\s*$", s):
|
|
863
|
+
in_notes = True
|
|
864
|
+
continue
|
|
865
|
+
if s.startswith("## "):
|
|
866
|
+
in_notes = False
|
|
867
|
+
if in_notes and notes_ok:
|
|
868
|
+
continue
|
|
869
|
+
leftovers.append(s)
|
|
870
|
+
if leftovers:
|
|
871
|
+
blockers.append("content outside the table with nowhere to go (%d line(s), first: %r) "
|
|
872
|
+
"-- project-wide notes belong in audit/%s"
|
|
873
|
+
% (len(leftovers), leftovers[0][:60], PROJECT_NOTES))
|
|
874
|
+
return blockers
|
|
875
|
+
|
|
876
|
+
|
|
730
877
|
def list_guides(root):
|
|
731
878
|
"""[(rel_to_ai_docs, path, meta, text)] for ai_docs/reference/GUIDE_*.md."""
|
|
732
879
|
ref = ai_path(root, "reference")
|
|
@@ -821,6 +968,30 @@ def cmd_index(root):
|
|
|
821
968
|
"a comprehension map (`source_kind: code`) -- see `guides.md`.\n",
|
|
822
969
|
encoding="utf-8")
|
|
823
970
|
print(f"[ok] guide router regenerated (empty stub): {gidx}")
|
|
971
|
+
return max(rc_registry(root), 0)
|
|
972
|
+
|
|
973
|
+
|
|
974
|
+
def rc_registry(root):
|
|
975
|
+
"""Write the generated workstream registry, or refuse and say why (F-028)."""
|
|
976
|
+
ws = list_workstreams(root)
|
|
977
|
+
if not ws:
|
|
978
|
+
return 0 # no sources: a hand-written handoff is untouched
|
|
979
|
+
blockers = registry_conversion_blockers(root)
|
|
980
|
+
hand = ai_path(root, "audit", "handoff.md")
|
|
981
|
+
if blockers:
|
|
982
|
+
print(f"[ERROR] {docs_dir()}/audit/handoff.md NOT regenerated -- it still holds "
|
|
983
|
+
"state no source accounts for:")
|
|
984
|
+
for b in blockers:
|
|
985
|
+
print(f" - {b}")
|
|
986
|
+
print(" Convert the whole registry at once (templates.md): converting "
|
|
987
|
+
"one row at a time is the state that loses the others.")
|
|
988
|
+
return 1
|
|
989
|
+
hand.parent.mkdir(parents=True, exist_ok=True)
|
|
990
|
+
hand.write_text(build_registry(root), encoding="utf-8")
|
|
991
|
+
print(f"[ok] workstream registry regenerated: {hand}")
|
|
992
|
+
if len(ws) > REGISTRY_CAP:
|
|
993
|
+
print(f"[warn] {len(ws)} open workstreams: the registry is meant to stay under "
|
|
994
|
+
f"{REGISTRY_CAP}. Nothing was truncated -- closing one is the fix.")
|
|
824
995
|
return 0
|
|
825
996
|
|
|
826
997
|
|
|
@@ -1257,8 +1428,33 @@ def cmd_validate(root, strict=False, hybrid=False):
|
|
|
1257
1428
|
elif norm_text(read_text(gidx)) != norm_text(build_guide_index(root)):
|
|
1258
1429
|
errors.append(f"{docs_dir()}/reference/INDEX.md not aligned with the guides: run 'sdlc_check.py index'")
|
|
1259
1430
|
|
|
1260
|
-
# Handoff: header and freshness
|
|
1431
|
+
# Handoff: alignment with its sources (F-028), then header and freshness
|
|
1261
1432
|
hand = ai / "audit" / "handoff.md"
|
|
1433
|
+
workstreams = list_workstreams(root)
|
|
1434
|
+
if workstreams:
|
|
1435
|
+
if not hand.is_file():
|
|
1436
|
+
errors.append(f"{docs_dir()}/audit/handoff.md missing while HANDOFF_*.md sources "
|
|
1437
|
+
"exist: the registry is the only place a cold agent sees the open "
|
|
1438
|
+
"workstreams -- run 'sdlc_check.py index'")
|
|
1439
|
+
elif norm_text(read_text(hand)) != norm_text(build_registry(root)):
|
|
1440
|
+
errors.append(f"{docs_dir()}/audit/handoff.md not aligned with its HANDOFF_*.md "
|
|
1441
|
+
"sources: run 'sdlc_check.py index'. A merge resolved by hand is "
|
|
1442
|
+
"exactly what this catches")
|
|
1443
|
+
if len(workstreams) > REGISTRY_CAP:
|
|
1444
|
+
warnings.append(f"{len(workstreams)} open workstreams: the registry is meant to "
|
|
1445
|
+
f"stay under {REGISTRY_CAP}")
|
|
1446
|
+
# Two files claiming one workstream is the collision this design does NOT
|
|
1447
|
+
# fix (two people opening the same work under different file names). It
|
|
1448
|
+
# would otherwise show up as two identical-looking rows and nothing else.
|
|
1449
|
+
seen = {}
|
|
1450
|
+
for p, meta in workstreams:
|
|
1451
|
+
wid = str(meta.get("workstream")).strip()
|
|
1452
|
+
if wid in seen:
|
|
1453
|
+
warnings.append(f"{docs_dir()}/audit/{p.name} and {seen[wid]} both claim "
|
|
1454
|
+
f"workstream '{wid}': the registry shows two rows for one "
|
|
1455
|
+
"workstream — decide which file owns it")
|
|
1456
|
+
else:
|
|
1457
|
+
seen[wid] = p.name
|
|
1262
1458
|
if hand.is_file():
|
|
1263
1459
|
m = re.search(r"(?:Date|Data):\s*(\d{4}-\d{2}-\d{2})", read_text(hand))
|
|
1264
1460
|
if not m:
|
|
@@ -83,7 +83,7 @@ Must-reads for this project, in order. The full manifest of canonical docs is
|
|
|
83
83
|
1. `reference/INDEX.md` — the guide router: which guide already governs the work you are about to do (generated).
|
|
84
84
|
2. `vision/project_vision.md` — why the project exists (check its Status first).
|
|
85
85
|
3. `strategic/architecture.md` — how it is built.
|
|
86
|
-
4. `audit/handoff.md` — where work stopped last session (
|
|
86
|
+
4. `audit/handoff.md` — where work stopped last session (generated from the `HANDOFF_*.md` beside it; never edited by hand).
|
|
87
87
|
|
|
88
88
|
Directory purposes: `vision/` (project direction), `strategic/` (architecture and
|
|
89
89
|
feature catalog), `reference/` (operative guides), `solutions/` (per-feature
|
|
@@ -389,36 +389,50 @@ States: PENDING (to analyze) | ANALYZED (analyzed, with reference) | SKIPPED (wi
|
|
|
389
389
|
| vendor/ | SKIPPED | - | vendored code |
|
|
390
390
|
```
|
|
391
391
|
|
|
392
|
-
## ai_docs/audit/handoff.md — the workstream registry
|
|
393
|
-
|
|
394
|
-
One row per OPEN workstream
|
|
395
|
-
|
|
396
|
-
|
|
397
|
-
|
|
398
|
-
|
|
399
|
-
|
|
400
|
-
|
|
401
|
-
|
|
402
|
-
|
|
403
|
-
|
|
404
|
-
|
|
405
|
-
|
|
406
|
-
|
|
407
|
-
|
|
408
|
-
|
|
409
|
-
|
|
392
|
+
## ai_docs/audit/handoff.md — the workstream registry (GENERATED)
|
|
393
|
+
|
|
394
|
+
One row per OPEN workstream. **Never written by hand**: `sdlc_check.py index` builds
|
|
395
|
+
it from the `HANDOFF_[topic].md` files, and `validate` errors when the two disagree.
|
|
396
|
+
It is an **inventory for lookup** (like the generated manifest), not a work board: no
|
|
397
|
+
assignment, no due dates, no execution ordering, no holder.
|
|
398
|
+
|
|
399
|
+
**Why generated, and not just one row per workstream.** Row-per-workstream alone was
|
|
400
|
+
tried and was not enough: two workstreams opened from one base still conflicted twice
|
|
401
|
+
in this file, because a file-global `Date:` header defeats row-level ownership no
|
|
402
|
+
matter how few rows each writer touches. So the truth moved into the per-workstream
|
|
403
|
+
file, and the header is derived (the newest `updated:` in the sources — a value, never
|
|
404
|
+
a filesystem timestamp, which git does not preserve). Two writers on two workstreams
|
|
405
|
+
now touch two different files. The generated view can still conflict at merge; that
|
|
406
|
+
conflict is resolved **mechanically** by re-running `index`, never by hand, and
|
|
407
|
+
`validate` refuses CLEAN until the file matches its sources.
|
|
408
|
+
|
|
409
|
+
Project-wide notes have their own source, `ai_docs/audit/project_notes.md`, appended
|
|
410
|
+
verbatim under `## Project-wide notes`. (Not `handoff_notes.md`: the `HANDOFF_*.md`
|
|
411
|
+
glob is case-insensitive on Windows and would collect it as a workstream.)
|
|
412
|
+
|
|
413
|
+
**Converting an existing project** — lazily, at the first write, and **all at once**.
|
|
414
|
+
Converting one row at a time is the state that loses the others: the next `index`
|
|
415
|
+
would regenerate from the one source and drop the rest. `index` refuses to write while
|
|
416
|
+
anything in the file is unaccounted for, and names it. A pre-1.17 narrative handoff
|
|
417
|
+
(`## Active features` / `## Next step` / `## Session notes`) is the same conversion:
|
|
418
|
+
each bullet becomes a `HANDOFF_[topic].md`, `## Session notes` becomes
|
|
419
|
+
`project_notes.md`. A project with no sources yet is not touched and reports nothing —
|
|
420
|
+
migrating a repository nobody is working on buys nothing.
|
|
410
421
|
|
|
411
422
|
```markdown
|
|
412
423
|
# Handoff — workstream registry
|
|
413
424
|
Date: 2026-06-11 (UTC)
|
|
414
425
|
|
|
426
|
+
<!-- GENERATED by sdlc_check.py index - do not edit by hand. Source of truth: the HANDOFF_*.md files in ai_docs/audit/. -->
|
|
427
|
+
|
|
415
428
|
| Workstream | Level | Branch | Status | Since | Next step | Details |
|
|
416
429
|
|---|---|---|---|---|---|---|
|
|
417
430
|
| F-001 SSO login | L3 | feature/sso-login | PROGRESS | 2026-06-10 | wire callback tests | HANDOFF_login_sso.md · ANALYSIS_login_sso.md |
|
|
418
|
-
| F-002 Audit refresh | L3 | feature/audit | PAUSED | 2026-06-02 | resume at Phase 4 |
|
|
431
|
+
| F-002 Audit refresh | L3 | feature/audit | PAUSED | 2026-06-02 | resume at Phase 4 | HANDOFF_audit_refresh.md · ANALYSIS_audit_refresh.md |
|
|
419
432
|
|
|
420
433
|
## Project-wide notes
|
|
421
|
-
|
|
434
|
+
|
|
435
|
+
<!-- from audit/project_notes.md: release pending, environment quirks that affect everyone -->
|
|
422
436
|
```
|
|
423
437
|
|
|
424
438
|
## ai_docs/audit/reviews/REVIEW_LOG.md
|
|
@@ -453,21 +467,47 @@ records the realization actually used — fresh subagent, one-shot client run, o
|
|
|
453
467
|
honest; writing nothing, or implying independence you did not have, is the failure
|
|
454
468
|
this column exists to prevent. `findings_real` is how many raised findings survived
|
|
455
469
|
triage: over time it is the only evidence of whether the gate earns its cost.
|
|
456
|
-
|
|
457
|
-
|
|
458
|
-
|
|
459
|
-
|
|
460
|
-
|
|
461
|
-
|
|
462
|
-
|
|
463
|
-
|
|
464
|
-
|
|
465
|
-
|
|
470
|
+
Concurrent reviews: `init` writes a `.gitattributes` stanza giving this file
|
|
471
|
+
`merge=union` — a **built-in** driver (no per-clone `git config`, unlike
|
|
472
|
+
`merge=ours`, which silently does nothing until every clone configures it).
|
|
473
|
+
Rows are date-stamped and their order carries no meaning, so a union merge keeps
|
|
474
|
+
both sides instead of asking a human to choose. It is defence in depth: without
|
|
475
|
+
git, or without the stanza, the outcome is today's — one conflict you resolve by
|
|
476
|
+
hand, never a lost row.
|
|
477
|
+
|
|
478
|
+
|
|
479
|
+
## ai_docs/audit/HANDOFF_[topic].md — one open workstream, its own file
|
|
480
|
+
|
|
481
|
+
**The authored home of that workstream's registry row**, and the only one: the
|
|
482
|
+
registry is generated from these files. **One exists for every OPEN workstream**, with
|
|
483
|
+
or without volatile state — a workstream whose file is missing has no row, and a
|
|
484
|
+
workstream with no row is invisible to the next cold agent. **DELETED at the feature's
|
|
485
|
+
closure**, in the same step that flips the ANALYSIS to COMPLETED: deleting it *is*
|
|
486
|
+
removing the row.
|
|
487
|
+
|
|
488
|
+
**The DRY boundary, restated because the file is no longer rare.** What used to keep
|
|
489
|
+
narrative out of it was that it barely existed; now it always does. So: the ANALYSIS
|
|
490
|
+
Diary keeps **what happened and why** (decisions, state of the work — survives
|
|
491
|
+
forever), and this file keeps **the row plus the resume logistics** — how to pick the work back up (branch,
|
|
492
|
+
worktree, uncommitted state, the next concrete command — worthless once resumed).
|
|
493
|
+
Prose that would still be worth reading after closure is in the wrong file, because
|
|
494
|
+
this one is deleted.
|
|
495
|
+
|
|
496
|
+
The frontmatter IS the row. `workstream:` is what marks the file as a source: without
|
|
497
|
+
it the file is still a perfectly good volatile note, and nothing generates.
|
|
466
498
|
|
|
467
499
|
```markdown
|
|
500
|
+
---
|
|
501
|
+
workstream: F-001 SSO login
|
|
502
|
+
level: L3
|
|
503
|
+
branch: feature/sso-login (worktree ../wt-sso)
|
|
504
|
+
status: PROGRESS
|
|
505
|
+
since: 2026-06-10
|
|
506
|
+
next: wire the callback tests
|
|
507
|
+
details: ANALYSIS_login_sso.md
|
|
508
|
+
updated: 2026-06-11
|
|
509
|
+
---
|
|
468
510
|
# HANDOFF: [topic] (ephemeral — deleted at closure)
|
|
469
|
-
Updated: 2026-06-11 (UTC)
|
|
470
|
-
Branch: feature/sso-login (worktree ../wt-sso)
|
|
471
511
|
|
|
472
512
|
## Resume state
|
|
473
513
|
<!-- uncommitted files, half-run migrations, env vars, running services -->
|
|
@@ -479,6 +519,19 @@ Branch: feature/sso-login (worktree ../wt-sso)
|
|
|
479
519
|
<!-- traps discovered this session that bite on resume (locks, CRLF, flaky test) -->
|
|
480
520
|
```
|
|
481
521
|
|
|
522
|
+
`updated:` is the date this file last changed, and the newest one across all sources
|
|
523
|
+
becomes the registry's `Date:` header — which is why no writer ever edits that header
|
|
524
|
+
and why two concurrent writers no longer collide on it. `details:` holds the *other*
|
|
525
|
+
pointers (the ANALYSIS, a review log entry); the generator prepends this file's own
|
|
526
|
+
name, so nothing points at itself by hand.
|
|
527
|
+
|
|
528
|
+
## ai_docs/audit/project_notes.md — the registry's project-wide notes (source)
|
|
529
|
+
|
|
530
|
+
Plain lines, no frontmatter, appended verbatim to the generated registry under
|
|
531
|
+
`## Project-wide notes`. Release pending, environment quirks, anything true for
|
|
532
|
+
everyone rather than for one workstream. It exists so that generating the registry
|
|
533
|
+
cannot destroy notes that belong to no workstream.
|
|
534
|
+
|
|
482
535
|
## ai_docs/strategic/architecture.md and existing_features.md
|
|
483
536
|
|
|
484
537
|
Canonical docs: they open with the header (`description:`/`status:`) so they enter the `INDEX.md` manifest cleanly.
|