@boshu2/learn
AOptionally analyze collections of durable verdicts for recurring evidence after the critical path. Triggers: "learn from verdicts", "mine validation history".
Install
agr install @boshu2/learn --target claudeWrites 1 file into .claude/skills/, pinned to git-282a5acf.
- .claude/skills/learn/SKILL.md
Document
name: learn description: 'Optionally analyze collections of durable verdicts for recurring evidence after the critical path. Triggers: "learn from verdicts", "mine validation history".' practices:
- continuous-learning
- evidence-based-engineering hexagonal_role: supporting consumes:
- verdict.v2 produces:
- learning-observations context_rel:
- kind: customer-of with: validate skill_api_version: 1 user-invocable: true metadata: graph_root: false tier: execution dependencies: [] capabilities: [analyze_verdict_collections] effects: [write_advisory_observations] canonical_status: canonical disposition: keep_off_path output_contract: advisory learning observations
Learn
Learn is an optional, off-path consumer of durable verdict.v2 collections.
It may summarize recurring evidence and propose a candidate deterministic check
for later human or caller evaluation.
Learn does not run during RPI, validate a subject, alter a verdict, mutate a plan, promote a rule, choose continuation, or mint lifecycle artifacts. Missing Learn output never changes whether a candidate is valid.
When invoked, bind every observation to verdict and finding digests, distinguish repeated objectives from repeated reviews of one objective, disclose the sample size, and stop at advisory evidence.
Overweight failures: a NOT_PROVEN or FAIL verdict carries more teaching
value than a PASS, because it names a rule the loop lacked. Harvest kernels
from failed lanes first — the canonical example is the mutating-check
quarantine in skills/validate/SKILL.md, a durable rule minted from a
NOT_PROVEN-then-PASS verdict pair.
Prune for provenance decay: every cited artifact must still resolve — the
file exists or the verdict digest is present under .agents/ao/verdicts/. A
citation that no longer resolves gets pruned rather than paraphrased, and
confidence in a lesson that has not been reproduced since its source decayed
goes down, not sideways.
When the caller asks for a durable artifact, write the observations under
.agents/scratch/learn/ and return the path; otherwise return them inline.
The write is advisory and TTL'd — it is never a source of record, and its
absence never changes whether a candidate is valid.
Trustgrade A
- passBody integrity
Whether the stored document is plausibly the kind of file the artifact declares, rather than something fetched by mistake.
- passType matchnot applicable to this artifact type
Whether the artifact is really the kind of thing its metadata claims it is.
- passFreshness
How long since the source repository was last pushed to.
- passPrompt injection
Scans the artifact's own text for instructions aimed at your agent rather than at you.
- passLicense
Whether the source repository declares an SPDX license permissive enough to redistribute.
How the grade is calculated
Each check contributes 0 points when it passes, 1 when it warns, and 2 when it fails. The total maps to a letter:
- Aevery check passed
- Bone warning
- Ctwo warnings
- Dprompt injection or body integrity failed, or three warnings
- Fone of those failed, and something else is wrong
These are automated hygiene checks, not a security audit, and not a dependency or vulnerability scan. A grade of A means nothing was flagged — not that the artifact is safe.
Versions
git-282a5acf46b02026-07-31