lessons-learned

name	lessons-learned
description	Learning repository that records successes, errors, and structured decisions to avoid repeating failures and feed the main-skill's memory system.
version	1.1.0
tags	["memory","governance","continuous-learning"]

Registration Structure

1. Context – task objective, involved skills, and expected impact.
2. Decision – choice made by main-skill, reasons, and considered alternatives.
3. Result – success, warning, or error; lessons learned.
4. Next steps – recommended adjustments or pending confirmations.

Operational Use

• Record every audit, significant error, or approved improvement suggestion.
• Consult whenever main-skill detects a task similar to a previously documented one.
• Keep links to logs (agf/logs/llm-errors.*) when there is a recurring error, to feed the memory system.

Improvement Suggestions

• Favor short and objective entries; use numbered lists for multi-factor decisions.
• Update the file immediately after critical actions to maintain the chain of trust.
• Use custom tags (e.g., #recurring-error, #model-success) if necessary to facilitate search.

2026-03-29 — Governance baseline automation

• Context: Connecting the deterministic agf/hooks/check-governance.sh baseline to real enforcement channels (git hooks, CI) so the governance claims are provably executed and observable.
• Decision: Installed a pre-commit hook that invokes the script and added .github/workflows/governance-check.yml that runs it on main push/PRs; documented the integrations in agf/hooks/README.md.
• Result: Local commits and CI now evaluate the same baseline before acceptance; no errors occurred while granting the hook executable status.
• Next steps: Monitor the workflow on future pushes; capture any anomalies in agf/logs/llm-errors.md and re-evaluate the threat model if enforcement levels change.

2026-03-29 — Governance hardening wrap-up

• Context: Completed the remaining TODO tasks for this session (repo init, DS_Store cleanup, Portuguese doc normalization, lessons-learned entry) to get the governance baseline fully compliant before moving on.
• Decision: Confirmed the repository links to upstream and uses .githooks/pre-commit plus CI workflow for governance checks, removed .DS_Store remnants, rewrote agf/docs/tech-notes.md in English, and noted the completion in agf/docs/TODO.md.
• Result: The governance baseline now runs cleanly with no locale-specific docs; the TODO backlog reflects the completed work; baseline automation is documented across README/hook index.
• Next steps: Keep documenting any new hardening effort here and capture user-facing decisions when working on follow-on tasks (e.g., lessons, threat models, or additional hooks).