principal-pi-skills 4.7.1 → 4.7.2

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.md CHANGED
@@ -132,27 +132,26 @@ assumptions-not-questions rule, the final-message-only rule. Anything outside a
132
132
  to every output.
133
133
 
134
134
  When maintaining this package, put lasting regressions in existing behavior/domain suites,
135
- not PR-named files. `npm test` checks current contracts/product behavior; historical receipts
136
- and one-off replay tooling belong under `tests/evidence/`, selected by `npm run verify:evidence`.
137
- Report those counts separately. Preserve archived evidence and its original candidate identity.
135
+ not PR-named files. `npm test` checks current contracts and product behavior.
136
+ Behavioural measurement—including the fidelity corpus and skill-harness evidence—lives in [principal-pi-skills-evals](https://github.com/mojomanyana/principal-pi-skills-evals); routing checks live here.
138
137
 
139
138
  ## Setup (pi)
140
139
 
141
- 1. `pi install git:github.com/mojomanyana/principal-pi-skills@v4.7.1` installs the eight
140
+ 1. `pi install git:github.com/mojomanyana/principal-pi-skills@v4.7.2` installs the eight
142
141
  skills, the four `/principal-*` commands, and the bootstrap
143
142
  extension, which loads automatically with the package. Install a tag, not a branch.
144
- 2. Subagents (optional): before npm publication, latest is 4.6.0 and npm @4.7.1 is not
143
+ 2. Subagents (optional): before npm publication, latest is 4.7.1 and npm @4.7.2 is not
145
144
  yet available. Use the installed tagged package's `scripts/install-agents.mjs` with Node
146
145
  (`install`, then `check`), locating its actual path rather than assuming a universal Pi
147
146
  install directory. Alternatively follow the concrete verified-tag disposable checkout
148
147
  recipe in [README Install](README.md#install-pi). After npm publication, pin both:
149
- `npx -p principal-pi-skills@4.7.1 principal-pi-agents install` and
150
- `npx -p principal-pi-skills@4.7.1 principal-pi-agents check`.
148
+ `npx -p principal-pi-skills@4.7.2 principal-pi-agents install` and
149
+ `npx -p principal-pi-skills@4.7.2 principal-pi-agents check`.
151
150
  The five definitions go to `${PI_CODING_AGENT_DIR:-~/.pi/agent}/agents`; foreign files
152
151
  are refused. Without them, the routing table still runs inline.
153
- Version 4.7.1 combines PR #58 and its repairs on the final merged commit. Existing v4.7.0
154
- at `448ac76` remains historical; leave it untouched. Tag creation and npm publication are
155
- separate: before the new tag exists, validate the actual candidate's skills and installer
152
+ Version 4.7.2 moves behavioral measurement to `principal-pi-skills-evals` while routing
153
+ checks remain here. Tag creation and npm publication are separate: before the new tag exists,
154
+ validate the actual candidate's skills and installer
156
155
  in isolation. Recorded model/evidence limitations remain; a release tag does not erase them.
157
156
  3. Context handoff (pi-daddy 0.33.0+): each skill's `allowed-tools` sets how much of this
158
157
  session a delegated child may receive. `architect`, `decide` and `plan` allow
package/CHANGELOG.md CHANGED
@@ -6,6 +6,21 @@ Where review revealed a prior claim or design decision didn't hold up under clos
6
6
 
7
7
  ---
8
8
 
9
+ ## [4.7.2] — 2026-10-01
10
+
11
+ **Changed — behavioral measurement moved out (#59).** The requirement-fidelity corpus,
12
+ skill-harness specifications and results, retained evidence, replay tooling, and their
13
+ repository tests now live in
14
+ [principal-pi-skills-evals](https://github.com/mojomanyana/principal-pi-skills-evals).
15
+ Routing inputs and baselines remain here, and `npm test` now covers only current product and
16
+ contract behavior; previously this repository also retained the behavioral corpus and
17
+ historical evidence checks.
18
+
19
+ **Reaffirmed — the 4.0.1 boundary.** Behavioral measurement belongs in the external
20
+ measurement repository rather than this runtime package. PR #59 applies that original 4.0.1
21
+ decision to the later requirement-fidelity material instead of maintaining a local exception.
22
+ No skill, agent, prompt, bootstrap, or generated contract text changed.
23
+
9
24
  ## [4.7.1] — 2026-10-01
10
25
 
11
26
  **Combined release candidate — all requirement-fidelity work and repairs (#58).** This
package/README.md CHANGED
@@ -125,7 +125,7 @@ bootstrap/BOOTSTRAP.md routing table + Next: vocabulary + model t
125
125
  extensions/bootstrap.ts pi extension: injects BOOTSTRAP.md at session start and after compaction
126
126
  scripts/ generator, installers, and checks behind `npm test`
127
127
  tests/{unit,install}/ current product/contract + clean-home install tests (node:test)
128
- tests/evidence/ opt-in offline receipt integrity and replay-tool checks
128
+ evals/{triggers.json,adversarial-triggers.json,baseline/} routing checks and baselines
129
129
  AGENTS.md routing + dispatch reference; the bootstrap injects its table automatically
130
130
  CHANGELOG.md release history
131
131
  ```
@@ -135,10 +135,10 @@ CHANGELOG.md release history
135
135
  1. **Skills + prompts** — install an immutable tag, not a branch:
136
136
 
137
137
  ```
138
- pi install git:github.com/mojomanyana/principal-pi-skills@v4.7.1
138
+ pi install git:github.com/mojomanyana/principal-pi-skills@v4.7.2
139
139
  ```
140
140
 
141
- The v4.7.1 `pi` manifest registers the eight skills, the four `/principal-*` commands,
141
+ The v4.7.2 `pi` manifest registers the eight skills, the four `/principal-*` commands,
142
142
  and the bootstrap extension — it loads automatically with
143
143
  the package; there is no separate extension-install step. Unpinned `main` moves under
144
144
  you, so install a tag if you want a fixed, nameable behavior.
@@ -148,8 +148,8 @@ CHANGELOG.md release history
148
148
  and reports success. `2.3.1` is the lowest safe version.
149
149
 
150
150
  2. **Subagents (optional).** Use the **same source** as the installed skills. Before npm
151
- publication, npm latest is 4.6.0; unpinned `npx` would install/check older definitions,
152
- and npm `@4.7.1` is not yet available. Either locate the actual installed tagged package
151
+ publication, npm latest is 4.7.1; unpinned `npx` would install/check older definitions,
152
+ and npm `@4.7.2` is not yet available. Either locate the actual installed tagged package
153
153
  (its path varies by Pi configuration) and run its `scripts/install-agents.mjs` with Node,
154
154
  or use this concrete matching-tag disposable checkout recipe:
155
155
 
@@ -158,8 +158,8 @@ CHANGELOG.md release history
158
158
  set -eu
159
159
  source_dir=$(mktemp -d)
160
160
  trap 'rm -rf -- "$source_dir"' EXIT
161
- git clone --depth 1 --branch v4.7.1 https://github.com/mojomanyana/principal-pi-skills.git "$source_dir/package"
162
- test "$(git -C "$source_dir/package" rev-parse HEAD)" = "$(git -C "$source_dir/package" rev-parse 'v4.7.1^{commit}')"
161
+ git clone --depth 1 --branch v4.7.2 https://github.com/mojomanyana/principal-pi-skills.git "$source_dir/package"
162
+ test "$(git -C "$source_dir/package" rev-parse HEAD)" = "$(git -C "$source_dir/package" rev-parse 'v4.7.2^{commit}')"
163
163
  git -C "$source_dir/package" rev-parse HEAD # retain the resolved source identity
164
164
  node "$source_dir/package/scripts/install-agents.mjs" install
165
165
  node "$source_dir/package/scripts/install-agents.mjs" check
@@ -169,14 +169,14 @@ CHANGELOG.md release history
169
169
  After npm publication, and only once the matching package is available, pin both commands:
170
170
 
171
171
  ```sh
172
- npx -p principal-pi-skills@4.7.1 principal-pi-agents install
173
- npx -p principal-pi-skills@4.7.1 principal-pi-agents check
172
+ npx -p principal-pi-skills@4.7.2 principal-pi-agents install
173
+ npx -p principal-pi-skills@4.7.2 principal-pi-agents check
174
174
  ```
175
175
 
176
- Version 4.7.1 combines PR #58 and its repairs under one new tag on the final merged
177
- commit. The existing v4.7.0 tag at `448ac76` remains historical and excludes later repairs;
178
- do not move it. Use the commands above after the v4.7.1 tag exists; npm publication is
179
- separate. The checkout check verifies tag/HEAD consistency, not independent tag trust;
176
+ Version 4.7.2 moves behavioral measurement to the separate
177
+ `principal-pi-skills-evals` repository while retaining routing checks here. Use the commands
178
+ above after the v4.7.2 tag exists; npm publication is separate. The checkout check verifies
179
+ tag/HEAD consistency, not independent tag trust;
180
180
  compare the recorded SHA with the release identity when provenance matters. Before tagging,
181
181
  validate the actual candidate's skills and installer in an isolated `PI_CODING_AGENT_DIR`.
182
182
  Known behavioral qualification limits remain documented; combining the release does not waive them.
@@ -232,17 +232,9 @@ CHANGELOG.md release history
232
232
 
233
233
  `npm test` remains the free gate: generated-contract drift, word budgets, frontmatter lint,
234
234
  installer and tarball behavior, `Next:` transition parity, and deterministic source-fidelity
235
- contract assertions and offline behavioral-corpus integrity/negative-oracle tests. These
236
- check current contracts and fixtures, not model behavior or instruction delivery.
237
-
238
- `npm run verify:evidence` separately runs `node --test tests/evidence/*.test.mjs`:
239
- offline historical receipt/hash integrity, retained failed observations, and fail-closed
240
- replay-parser checks. It never invokes Pi or a model; the parser check skips with a reason
241
- if optional Python is absent. It is not part of `npm test` or default CI. Passing it does
242
- not qualify current model behavior or restamp an archived candidate. Historical test
243
- totals (including 141 unit tests) describe their recorded runs, not today's suite size.
244
- Archived manifests retain their original candidate identity even when the working tree
245
- has changed; reports, raw traces and manifests are not rewritten by test cleanup.
235
+ contract assertions. These check current product contracts, not model behavior or instruction delivery.
236
+
237
+ Behavioural measurement—including the requirement-fidelity corpus and skill-harness evidence—lives in [principal-pi-skills-evals](https://github.com/mojomanyana/principal-pi-skills-evals); routing checks live here.
246
238
 
247
239
  **Enforcement boundary:** Build/Review instructions ask the model to organize regressions
248
240
  by lasting behavior, not PR/finding IDs, and distinguish product coverage from historical
@@ -255,47 +247,6 @@ The package commands and CI mechanically select suites; pi-daddy's delegated too
255
247
  control available capabilities, not test names or correctness. Refresh installed skill/agent
256
248
  definitions when adopting a revision: editing this checkout does not update another installed copy.
257
249
 
258
- A narrow local opt-in behavioral regression corpus is authorized for requirement fidelity,
259
- superseding only the blanket removal of focused harness fixtures below. It is **partially measured, not fully qualified**:
260
- 46 runnable component cases now live in `<skill>/tests/specification.yaml` for Plan, Build,
261
- Review, Investigate, Decide, Architect and Debug. Shared synthetic fixtures, 51-row long-source
262
- oracle, all F01–F27 status/replay recipes and runner limitations are documented in
263
- [evals/requirement-fidelity/README.md](evals/requirement-fidelity/README.md), with an
264
- [evidence index](evals/requirement-fidelity/evidence.md). The three evidence-scope cases have
265
- [bounded live observations](evals/requirement-fidelity/evidence/evidence-scope/README.md):
266
- F25/F27 raw PASS; F26's old reporting failure is retained, with a
267
- [new corrected-rubric PASS and separate execution audit](evals/requirement-fidelity/evidence/regression-evidence-alignment/README.md).
268
- [Whole-PR repairs](evals/requirement-fidelity/evidence/full-review-repairs/README.md)
269
- retain the earlier report-safety observations and 5/9 run.
270
- [Latest Build handoff validation](evals/requirement-fidelity/evidence/build-handoff/README.md)
271
- is **4/9 NOT READY**; three fully captured cases verify saved patches and functional regressions
272
- but retain caveat-delivery, persistence-stop and reproducibility limitations.
273
- The earlier narrow direct audit remains UNVERIFIED with explicit manual inspection.
274
- Older 43-case results remain historical.
275
- Retained live runs include failures;
276
- a manually orchestrated source→plan→build→review chain, full long-plan inspection, and two
277
- normal skill-loading probes provide narrower positive evidence, not an all-case pass.
278
- PR #58 retains three direct blocking runs of the unchanged critical F04 negative and three
279
- successful supplied-definition companions. Actual registered workflow prompts demonstrate
280
- repeat-report preservation, absent-ignore initialization and an approval-gated inline bugfix.
281
- These are bounded observations, not passing harness grades; fresh-session resumed repair and
282
- delegated transport remain unmeasured. Three parent-only replay recipes track partial coverage.
283
- Retain results.yaml, synthetic transcripts/traces, changed-file diffs
284
- and complete output artifacts under `<skill>/tests/results/<runner-model>/<run>/`; use narrow
285
- ignore rules for incidental caches/HTML only. Review evidence for secrets before staging.
286
- No model calls in `npm test`, no harness runtime dependency, no fixtures/results in the npm
287
- package. Broad comparisons and qualification remain external.
288
-
289
- Record candidate/spec/fixture identities, runner/Pi versions, subject/judge, mode, repetitions,
290
- commands and retained artifacts. Separate forced instruction following from normal runtime
291
- loading; missing delivery/retention evidence is UNKNOWN, never PASS. Installed skill-harness
292
- 0.21.0 preflight exposed schema-2 and retention gaps; no compatible published update is
293
- available per the implementation preflight. P0/schema-3/full behavioral validation has not
294
- passed. An unsupported Spark provider response prevented lower-cost robustness measurements;
295
- CLI exit zero was not counted as success. Remaining Build/Architect failures, tool-ceiling
296
- violations, and unmeasured combinations are recorded rather than waived. No prior deleted
297
- 4/4 claim is restored. Focused results establish only their recorded candidate/model/delivery cells.
298
-
299
250
  Two opt-in routing checks use only the eight authored frontmatter descriptions. Run
300
251
  `npm run check:routing-collisions` for all 56 directed description pairs and
301
252
  `npm run check:routing-triggers` for the three-run synthetic trigger suite. Both require
@@ -328,7 +279,7 @@ Decisions taken for 4.0, all closed:
328
279
  | Build delegation | `principal-build` agent; inline when there is no multi-step plan file or no subagent tool |
329
280
  | Plan persistence | multi-step plans to git-ignored `.principal/plans/<slug>.md`; no date prefix because plan has no clock, and resume matches on the `## Plan:` line |
330
281
  | Decide vs architect | both kept; decide answers "should we / which", architect answers "how is it structured" |
331
- | Measurement | broad qualification stays external; narrow opt-in requirement-fidelity regressions permitted locally (see Validation), partially measured but not fully qualified |
282
+ | Measurement | behavioural measurement lives in [principal-pi-skills-evals](https://github.com/mojomanyana/principal-pi-skills-evals); routing checks stay here |
332
283
 
333
284
  Two implementation notes that differ from the obvious reading: the bootstrap is injected as
334
285
  a user-role message wrapped in `<IMPORTANT>`, because pi's `context` hook can only insert
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "principal-pi-skills",
3
- "version": "4.7.1",
3
+ "version": "4.7.2",
4
4
  "description": "Skill framework for principal-level software engineering: eight skills (decide, architect, plan, build, review, debug, investigate, git-ops), a pi bootstrap, and two approval-gated workflows.",
5
5
  "license": "MIT",
6
6
  "author": "Nemanja Alavanja",
@@ -22,7 +22,6 @@
22
22
  },
23
23
  "scripts": {
24
24
  "test": "npm run generate:check && npm run test:unit && npm run test:install && npm run check:pack && npm run lint:skills",
25
- "verify:evidence": "node --test tests/evidence/*.test.mjs",
26
25
  "test:unit": "node --test tests/unit/*.test.mjs && node scripts/check-word-budgets.mjs",
27
26
  "generate": "node scripts/generate-contracts.mjs",
28
27
  "generate:check": "node scripts/generate-contracts.mjs --check",