mandrel 2.79.0 → 2.81.0
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/.agents/agents/auditor.md +8 -3
- package/.agents/audit-checklists/data-model.md +9 -1
- package/.agents/audit-checklists/security.md +15 -1
- package/.agents/docs/agentrc-reference.json +4 -0
- package/.agents/docs/configuration.md +4 -3
- package/.agents/docs/execution-reference.md +25 -5
- package/.agents/docs/workflows.md +2 -2
- package/.agents/schemas/agentrc.schema.json +31 -3
- package/.agents/schemas/audit-rules.json +15 -2
- package/.agents/scripts/deliver-run.js +43 -8
- package/.agents/scripts/lib/audit-to-stories/seed-from-findings.js +37 -2
- package/.agents/scripts/lib/baselines/duplication-scanner.js +7 -1
- package/.agents/scripts/lib/baselines/refresh-service.js +6 -0
- package/.agents/scripts/lib/cli-args.js +16 -2
- package/.agents/scripts/lib/config/explain.js +1 -1
- package/.agents/scripts/lib/config/gates/duplication.schema.js +10 -2
- package/.agents/scripts/lib/config-settings-schema.js +18 -1
- package/.agents/scripts/lib/findings/audit-finding-record.js +95 -0
- package/.agents/scripts/lib/findings/route-finding.js +4 -3
- package/.agents/scripts/lib/generated/agentrc-validator.js +1 -1
- package/.agents/scripts/lib/observability/close-telemetry.js +17 -2
- package/.agents/scripts/lib/orchestration/complexity-gate.js +58 -11
- package/.agents/scripts/lib/orchestration/plan-persist/audit-attribution.js +182 -0
- package/.agents/scripts/lib/orchestration/plan-persist/audit-provenance.js +11 -10
- package/.agents/scripts/lib/orchestration/plan-persist/collision-message.js +52 -0
- package/.agents/scripts/lib/orchestration/plan-persist/run-plan-persist.js +56 -9
- package/.agents/scripts/lib/orchestration/plan-persist/source-rewrite.js +177 -0
- package/.agents/scripts/lib/orchestration/plan-persist/story-ops.js +202 -57
- package/.agents/scripts/lib/orchestration/plan-persist/supersede-ops.js +57 -15
- package/.agents/scripts/lib/orchestration/plan-persist/wave-collision-gate.js +8 -4
- package/.agents/scripts/lib/orchestration/plan-persist/wave-serialisation.js +3 -2
- package/.agents/scripts/lib/orchestration/resolve-stories.js +12 -3
- package/.agents/scripts/lib/orchestration/resume-effort.js +121 -0
- package/.agents/scripts/lib/orchestration/single-story-close/phases/options.js +13 -2
- package/.agents/scripts/lib/orchestration/single-story-close/runner.js +10 -2
- package/.agents/scripts/lib/orchestration/ticketing/reads.js +3 -0
- package/.agents/scripts/lib/qa/resolve-qa-contract.js +28 -9
- package/.agents/scripts/lib/templates/decomposer-prompts.js +2 -2
- package/.agents/scripts/lib/wave-runner/footprint.js +114 -17
- package/.agents/scripts/lib/wave-runner/ready-set.js +16 -5
- package/.agents/scripts/plan-persist.js +20 -6
- package/.agents/scripts/providers/github/tickets.js +2 -1
- package/.agents/scripts/resolve-stories.js +8 -2
- package/.agents/scripts/single-story-close.js +11 -3
- package/.agents/workflows/audit-data-model.md +94 -10
- package/.agents/workflows/audit-security.md +136 -49
- package/.agents/workflows/audit-to-stories.md +22 -7
- package/.agents/workflows/helpers/audit-lens-core.md +8 -5
- package/.agents/workflows/helpers/deliver-reference.md +7 -2
- package/.agents/workflows/helpers/deliver-story.md +5 -3
- package/.agents/workflows/helpers/plan-reference.md +53 -5
- package/.agents/workflows/mandrel-deliver.md +9 -4
- package/.agents/workflows/mandrel-plan.md +7 -1
- package/docs/CHANGELOG.md +25 -0
- package/package.json +2 -1
|
@@ -238,7 +238,14 @@ work this plan already tracked.
|
|
|
238
238
|
| --- | --- |
|
|
239
239
|
| Present | **Exactly** the identities listed — siblings' groups never leak in. |
|
|
240
240
|
| Present but empty (`{}`) | Nothing. "Owns no findings" is a real answer. |
|
|
241
|
-
| **Absent
|
|
241
|
+
| **Absent**, N>1 plan, seed carries `audit-finding` records | The findings whose files this Story's `changes[]` covers (exact path or glob; the primary file first, the others only when no Story covers it), plus every **unattributed** finding. |
|
|
242
|
+
| **Absent**, otherwise (N==1, or a record-less seed) | The **whole seed's** footers (the union) — the recall-safe default. |
|
|
243
|
+
|
|
244
|
+
An unattributed finding — one no Story owns by `provenance` or by path — is
|
|
245
|
+
stamped on every Story without an authored `provenance` and listed in
|
|
246
|
+
persist's warnings (`unattributed audit finding <sha12> (<file>) — carried on
|
|
247
|
+
all N Stories`). `audit::*` labels follow the stamped findings under
|
|
248
|
+
attribution, and the seed union otherwise.
|
|
242
249
|
|
|
243
250
|
**The union fallback is load-bearing, not legacy** — hand-carrying provenance
|
|
244
251
|
out of the seed's HTML comments is a step an author forgets. Attribution is
|
|
@@ -251,8 +258,9 @@ owning Story has **closed** is masked by an open neighbour. With ownership
|
|
|
251
258
|
stamped, the issue carrying a finding's own fingerprint decides both the match
|
|
252
259
|
and its state (`lib/findings/route-finding.js`).
|
|
253
260
|
|
|
254
|
-
The audit path
|
|
255
|
-
|
|
261
|
+
The audit path attributes mechanically from the per-finding `audit-finding`
|
|
262
|
+
records the seed carries beside its per-group footers — see
|
|
263
|
+
[`audit-to-stories`](../audit-to-stories.md). A `--seed`
|
|
256
264
|
or `--tickets` plan has nothing to attribute and omits the field.
|
|
257
265
|
|
|
258
266
|
## Cross-Story conflict analysis at persist
|
|
@@ -546,10 +554,50 @@ Whichever channel supplies them, the supersede-map partition above still
|
|
|
546
554
|
fail-closes: a `--tickets` run whose Stories forgot `supersedes[]` is
|
|
547
555
|
**caught** (`source ticket #N is not claimed by any Story`).
|
|
548
556
|
|
|
557
|
+
## Rewriting a one-for-one source in place
|
|
558
|
+
|
|
559
|
+
When a tickets-mode plan is **one-for-one** — exactly one source ticket and
|
|
560
|
+
exactly one drafted Story — persist rewrites that ticket instead of filing a
|
|
561
|
+
new Story and closing the source as its duplicate. The issue number, its
|
|
562
|
+
comment history and its non-`agent::*` labels survive. This is the default.
|
|
563
|
+
Gate #2 still asks, so the operator can choose either way:
|
|
564
|
+
|
|
565
|
+
| Choice | What persist does |
|
|
566
|
+
| --- | --- |
|
|
567
|
+
| **Rewrite #N in place** (Recommended) | No flag — the default. |
|
|
568
|
+
| **File a new Story and close #N** | `--no-rewrite-source` |
|
|
569
|
+
|
|
570
|
+
Under `--yes` the question is skipped and the default applies.
|
|
571
|
+
|
|
572
|
+
**The write order** keeps "`agent::ready` means fully persisted" true even
|
|
573
|
+
for a source that was already `agent::ready`:
|
|
574
|
+
|
|
575
|
+
1. If the source carries no `plan-rewrite` comment, post one holding its
|
|
576
|
+
original title and body in a collapsed block. **First write wins** — a
|
|
577
|
+
later re-plan never replaces it, so the pre-plan evidence survives.
|
|
578
|
+
2. One PATCH sets the new title and body (with the plan fingerprint, so a
|
|
579
|
+
retried persist adopts the rewritten issue instead of duplicating it),
|
|
580
|
+
adds the Story's labels and this run's `plan-run::` label, and removes
|
|
581
|
+
every `agent::*` label and any earlier run's `plan-run::` label.
|
|
582
|
+
3. The `story-plan-state` comment, then the terminal `agent::ready` flip.
|
|
583
|
+
|
|
584
|
+
The supersede phase lists the rewritten id under `supersede.rewritten` and
|
|
585
|
+
neither comments on nor closes it; the result's `rewrite` field names the
|
|
586
|
+
ticket or why there was none.
|
|
587
|
+
|
|
588
|
+
**Falls back to create-and-close** with no warning for any other shape
|
|
589
|
+
(several sources, several Stories), under `--no-rewrite-source`, and under
|
|
590
|
+
`--no-close-superseded` — that flag means the remainder of the source must
|
|
591
|
+
stay open as itself, which a rewrite would destroy. It **falls back with a
|
|
592
|
+
warning** naming the reason when the source is closed, inaccessible, a
|
|
593
|
+
`type::epic` container, or mid-delivery (`agent::executing`,
|
|
594
|
+
`agent::closing`, or `agent::blocked` — a paused delivery still owns its
|
|
595
|
+
branch and PR).
|
|
596
|
+
|
|
549
597
|
## Closing superseded source tickets
|
|
550
598
|
|
|
551
|
-
**Default on
|
|
552
|
-
issue naming the specific Story that claims it — plus that Story's optional
|
|
599
|
+
**Default on** for every source not rewritten in place (above). After the
|
|
600
|
+
Stories exist, persist comments on each source issue naming the specific Story that claims it — plus that Story's optional
|
|
553
601
|
per-supersede `note` — and closes it with reason **`duplicate`**
|
|
554
602
|
(`state_reason`): the claiming Story carries the work, and nothing has
|
|
555
603
|
shipped at persist time, so `completed` would be a lie. A provider that
|
|
@@ -32,7 +32,7 @@ you read:
|
|
|
32
32
|
| Invocation | Shape | Behavior |
|
|
33
33
|
| --- | --- | --- |
|
|
34
34
|
| `/mandrel-deliver` | bare | List the open `agent::ready` Stories and ask which to deliver. Deliver nothing until answered. |
|
|
35
|
-
| `/mandrel-deliver 4712` | ids | One Story via `helpers/deliver-story.md`, **inline in this session** — no `story-worker` spawn. |
|
|
35
|
+
| `/mandrel-deliver 4712` | ids | One Story via `helpers/deliver-story.md`, **inline in this session** — no `story-worker` spawn. A resume of a Story that blocked before dispatches one worker instead (step 3). |
|
|
36
36
|
| `/mandrel-deliver 4712 4713 …` | ids | Resolve the set, then beat `deliver-run.js` — it sequences by the discovered graph and hands back the spawns and closes. |
|
|
37
37
|
| `/mandrel-deliver 4712 - 4716` | ids | A **range** — every id in the inclusive span. |
|
|
38
38
|
| `/mandrel-deliver 4700` (a `type::epic`) | ids | The Epic's **open** child Stories. Mixes with Story ids. |
|
|
@@ -70,7 +70,9 @@ to an attended run.
|
|
|
70
70
|
present the order in step 2, but do **not** thread them into step 3 — every
|
|
71
71
|
beat re-resolves the graph. A one-id run comes back `dispatchMode: "inline"`
|
|
72
72
|
and runs [`helpers/deliver-story.md`](helpers/deliver-story.md) in this
|
|
73
|
-
session; step 3 is the multi-Story path.
|
|
73
|
+
session; step 3 is the multi-Story path. A one-id run that comes back
|
|
74
|
+
`dispatchMode: "subagent"` is a resume of a Story that blocked before: take
|
|
75
|
+
step 3 with that one id, so its worker can carry the resume effort. An Epic id expands to its open child
|
|
74
76
|
Stories first — **announce it**. It hard-errors (exit 1) on an id that is
|
|
75
77
|
neither a Story nor an Epic, on an Epic with no open children, or on edges
|
|
76
78
|
it cannot read — a missing gate would co-dispatch against an unlanded
|
|
@@ -96,8 +98,11 @@ to an attended run.
|
|
|
96
98
|
maintained across beats by you**. Cross-run de-confliction via the assignee
|
|
97
99
|
lease is automatic. Branch on the exit code:
|
|
98
100
|
- **0** — spawn one `story-worker` per `ready[]` entry, **all in one turn**,
|
|
99
|
-
each with that entry's `promptPath` file as its prompt
|
|
100
|
-
|
|
101
|
+
each with that entry's `promptPath` file as its prompt. When the entry's
|
|
102
|
+
`effort` is non-null, pass it as that Agent call's `effort`; when it is
|
|
103
|
+
`null`, omit the parameter so the worker inherits the session. Run each
|
|
104
|
+
`close[]` entry's command foreground and serialized as hand-offs arrive.
|
|
105
|
+
An empty
|
|
101
106
|
`ready` with work in flight means "waiting", so beat again — **unless
|
|
102
107
|
`stalledDispatch[]` is non-empty**: those ids were dispatched but still
|
|
103
108
|
read `agent::ready`, which is a live init window the first beats or two
|
|
@@ -121,6 +121,11 @@ than one Story** (a split always earns operator eyes, `--force-review` or
|
|
|
121
121
|
not), or when the operator asked to review (`--force-review`).
|
|
122
122
|
Under `--yes`, auto-proceed.
|
|
123
123
|
|
|
124
|
+
**One-for-one (tickets mode, one source → one Story)** — ask with
|
|
125
|
+
AskUserQuestion: **Rewrite #N in place (Recommended)** or **File a new Story
|
|
126
|
+
and close #N** (`--no-rewrite-source`). Under `--yes`, rewrite
|
|
127
|
+
([ref](helpers/plan-reference.md)).
|
|
128
|
+
|
|
124
129
|
**Gate #3 — adopt, else create.** Offer the top `epicCandidates[]` Epic at
|
|
125
130
|
**any N** (`--epic <id>`); else, at **N>2**, a new container (`--epic-title` /
|
|
126
131
|
`--epic-goal`). Never unasked ([ref](helpers/plan-reference.md)).
|
|
@@ -151,7 +156,8 @@ Persist creates `type::story` issue(s), a **metadata-only** `plan-run::<id>`
|
|
|
151
156
|
label, `blocked by #<id>` footers for every `depends_on` edge, and on a Gate #3
|
|
152
157
|
yes links them under a `type::epic` container (adopted or new). `agent::ready`
|
|
153
158
|
is the **terminal** flip after receipts land; stdout is pure JSON. Tickets mode
|
|
154
|
-
|
|
159
|
+
rewrites a one-for-one source in place, else comments on and closes each
|
|
160
|
+
source id ([ref](helpers/plan-reference.md)).
|
|
155
161
|
|
|
156
162
|
## Constraints
|
|
157
163
|
|
package/docs/CHANGELOG.md
CHANGED
|
@@ -15,6 +15,31 @@ All notable changes to this project will be documented in this file.
|
|
|
15
15
|
-->
|
|
16
16
|
<!-- markdownlint-disable-file MD004 MD012 MD037 -->
|
|
17
17
|
|
|
18
|
+
## [2.81.0](https://github.com/dsj1984/mandrel/compare/mandrel-v2.80.0...mandrel-v2.81.0) (2026-10-11)
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
### Added
|
|
22
|
+
|
|
23
|
+
* /mandrel-plan rewrites a one-for-one source ticket in place instead of filing a new Story ([#5604](https://github.com/dsj1984/mandrel/issues/5604)) ([#5610](https://github.com/dsj1984/mandrel/issues/5610)) ([5e0f621](https://github.com/dsj1984/mandrel/commit/5e0f6215680ee9d9fec079a93fd51d531d59023a))
|
|
24
|
+
* agent cost report: per-subagent effort, model, tokens and wall time from Claude Code transcripts ([#5607](https://github.com/dsj1984/mandrel/issues/5607)) ([#5611](https://github.com/dsj1984/mandrel/issues/5611)) ([c1680ca](https://github.com/dsj1984/mandrel/commit/c1680ca12298e21a51d8000437a675fb94bd23f2))
|
|
25
|
+
* audit-security: trust-boundary attack classes, a Boundary candidate-gate field, and lens-specific severity anchors (adapted from Cloudflare's MIT security-audit skill) ([#5606](https://github.com/dsj1984/mandrel/issues/5606)) ([#5612](https://github.com/dsj1984/mandrel/issues/5612)) ([34dc807](https://github.com/dsj1984/mandrel/commit/34dc8077f7d2392385355519180d6f0122f27770))
|
|
26
|
+
* duplication gate now scans TS/TSX/JSX; re-seat via duplication:reanchor ([#5605](https://github.com/dsj1984/mandrel/issues/5605)) ([#5609](https://github.com/dsj1984/mandrel/issues/5609)) ([5f73168](https://github.com/dsj1984/mandrel/commit/5f731680c4c16aa1ba11b1d2f02032af931ea5aa))
|
|
27
|
+
* escalate effort per call on a blocked-Story resume instead of raising whole-session effort ([#5608](https://github.com/dsj1984/mandrel/issues/5608)) ([#5617](https://github.com/dsj1984/mandrel/issues/5617)) ([4285cdf](https://github.com/dsj1984/mandrel/commit/4285cdf55b057723f60dbe17808c23e421f645d7))
|
|
28
|
+
|
|
29
|
+
|
|
30
|
+
### Fixed
|
|
31
|
+
|
|
32
|
+
* **test:** attribute temp-paths leak writes by nonce, not stream growth (refs [#5615](https://github.com/dsj1984/mandrel/issues/5615)) ([#5616](https://github.com/dsj1984/mandrel/issues/5616)) ([0f33cec](https://github.com/dsj1984/mandrel/commit/0f33cecfe4db122074d79516b2d0c70c3e92bd0f))
|
|
33
|
+
|
|
34
|
+
## [2.80.0](https://github.com/dsj1984/mandrel/compare/mandrel-v2.79.0...mandrel-v2.80.0) (2026-10-09)
|
|
35
|
+
|
|
36
|
+
|
|
37
|
+
### Added
|
|
38
|
+
|
|
39
|
+
* audit-data-model: Dead Schema dimension and opt-in read-only live-schema verification pass ([#5594](https://github.com/dsj1984/mandrel/issues/5594)) ([#5595](https://github.com/dsj1984/mandrel/issues/5595)) ([3738dbe](https://github.com/dsj1984/mandrel/commit/3738dbe4af5bb7d71b97701684433d1feb13b839))
|
|
40
|
+
* audit-seeded N>1 plans: attribute each finding's footers and audit:: label to the Stories that cover it ([#5597](https://github.com/dsj1984/mandrel/issues/5597)) ([#5600](https://github.com/dsj1984/mandrel/issues/5600)) ([a8529be](https://github.com/dsj1984/mandrel/commit/a8529bee4f51771677d646275081df6deb3a6543))
|
|
41
|
+
* collision predicate: a glob collides only with paths it can cover, and the message names the declarer ([#5596](https://github.com/dsj1984/mandrel/issues/5596)) ([#5599](https://github.com/dsj1984/mandrel/issues/5599)) ([dfea48c](https://github.com/dsj1984/mandrel/commit/dfea48caf755577e625002bb6873e2d54291e045))
|
|
42
|
+
|
|
18
43
|
## [2.79.0](https://github.com/dsj1984/mandrel/compare/mandrel-v2.78.0...mandrel-v2.79.0) (2026-10-09)
|
|
19
44
|
|
|
20
45
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "mandrel",
|
|
3
|
-
"version": "2.
|
|
3
|
+
"version": "2.81.0",
|
|
4
4
|
"description": "Claude Code-first opinionated workflow framework: instructions, skills, rules, and SDLC workflows that govern AI coding assistants.",
|
|
5
5
|
"files": [
|
|
6
6
|
".agents/",
|
|
@@ -21,6 +21,7 @@
|
|
|
21
21
|
"test:integration": "node .agents/scripts/run-tests.js --tier integration",
|
|
22
22
|
"test:e2e": "node .agents/scripts/run-tests.js --tier e2e",
|
|
23
23
|
"test:profile": "node scripts/run-test-profile.js",
|
|
24
|
+
"report:agent-cost": "node scripts/agent-cost-report.js",
|
|
24
25
|
"test:census": "SPAWN_CENSUS_OUT=\"${SPAWN_CENSUS_OUT:-temp/census.json}\" NODE_OPTIONS=\"--require $PWD/tests/fixtures/spawn-census.cjs\" node .agents/scripts/run-tests.js",
|
|
25
26
|
"test:isolate": "node scripts/test-isolate.js",
|
|
26
27
|
"verify": "node scripts/run-verify.js",
|