feat: add typed experiment ledger

[divo12]

Summary

• add strict immutable experiment-attempt contracts and a record_experiment MCP tool
• persist idempotent, bounded artifacts under .workspace/experiments/ while reusing the hardened local artifact publisher
• require newly prepared agent programs to ledger every terminal admit/reject decision

Verification

• uv run ruff format --check on all changed Python files
• uv run ruff check src tests plugins/openflywheel/scripts/mcp_server.py
• uv run mypy src tests plugins/openflywheel/scripts/mcp_server.py
• uv run pytest --cov=ofw --cov-report=term-missing --cov-fail-under=90 -q — 187 passed, 2 skipped, 95.40% coverage
• uvx --from radon radon cc -s -a src tests plugins/openflywheel/scripts/mcp_server.py — average A; changed ledger functions at 5 or below
• all three skill validators and plugin validation

Scope and follow-ups

• existing prepared workspaces keep their generated PROGRAM.md; the ledger instruction applies to newly prepared workspaces
• listing, summaries, dashboards, and search-tree scheduling remain out of scope

---

Summary by cubic

Previously, terminal experiment decisions were not persisted. This adds a typed, immutable, idempotent ledger under .workspace/experiments/ and requires newly prepared agent programs to call record_experiment; existing prepared workspaces keep their generated PROGRAM.md.

Changes

• Admitted attempts require verifier receipts, cost, and latency; rejected attempts require a reason and may omit unavailable evidence.
• Artifacts store references and aggregate measurements only, never Langfuse trace payloads.
• Reuses a hardened local artifact publisher for failure and experiment artifacts with bounded writes, workspace checks, and conflict detection.
• Adds strict typed contracts and fail-closed validation for attempts and MCP input.
• Agents stop if recording fails instead of committing an unledgered change.

^{Written for commit 12244ef. Summary will update on new commits.}

[cubic-dev-ai]

1 issue found across 15 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="plugins/openflywheel/.codex-plugin/plugin.json">

<violation number="1" location="plugins/openflywheel/.codex-plugin/plugin.json:14">
P1: Installed plugin instances still launch an MCP commit that does not expose `record_experiment`, so this new description advertises an unavailable feature. Update `.mcp.json` to pin a release containing the ledger tool.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>