feat: the lessons loop, and six lessons approved and published

Capture, distil, review, publish, retire (ADR-0004). Lesson candidates of
past Executions are grouped into what recurred; a lesson-curator role,
outside any Execution, proposes a lesson per group worth one; each proposal
is an issue file a person approves or rejects; an approved lesson is
written into the provider's lessons Skill, which a Specialist is given only
once it holds lessons for that Specialist's Deliverable. LESSONS_ENABLED
off runs without them, for an evaluation with and without.

The curator proposed six lessons from 21 Executions. The owner approved all
six on 2026-10-09 and they are published: five for the target architect,
one for the IaC author. None is proved by an evaluation yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-10-09 15:19:32 +01:00
parent 59a641fded
commit a7bc5310f3
29 changed files with 1560 additions and 6 deletions

17
app/learning/__init__.py Normal file
View File

@@ -0,0 +1,17 @@
"""The lessons loop (ADR-0004): what Executions got wrong becomes reviewed Skill changes.
capture each Execution records lesson candidates app/states/lessons.py
distil recurring candidates are grouped, and a curator
proposes a lesson for each group worth one evidence.py, curator.py
review each proposal is an issue file a person decides proposals.py
publish an approved lesson is written into the lessons
Skill the provider bundle assigns to roles skill.py
prove the evaluation runs with and without lessons
retire a lesson that did not help, or that the upstream
Skills now cover, is removed skill.py
Nothing here runs inside an Execution, and no lesson reaches an agent without a
person's approval. Lesson candidates are client data: they are read from the
Executions' records and never copied into the repository; a lesson and its issue
file carry counts and a general description only.
"""