New /pre-ci & /feedback commands for Claude

dev/commands/feedback.md

@@ -0,0 +1,92 @@
+# Session Feedback
+
+Analyze this conversation and identify what documentation or context was missing, incomplete, or incorrect. The goal is to continuously improve the project's knowledge base so future conversations are more efficient.
+
+## Step 1: Session Analysis
+
+Reflect on the work done in this conversation. For each area, identify friction points:
+
+1. **Exploration overhead**: What parts of the codebase did you have to discover by searching that should have been documented? (e.g., patterns, conventions, module responsibilities)
+2. **Wrong assumptions**: Did you make incorrect assumptions due to missing or misleading documentation?
+3. **Repeated patterns**: Did you discover recurring patterns or conventions that aren't documented anywhere?
+4. **Missing context**: What background knowledge would have helped you start faster? (e.g., architecture decisions, data flow, naming conventions)
+5. **Tooling gaps**: Were there commands, scripts, or workflows that you had to figure out?
+
+## Step 2: Documentation Audit
+
+For each friction point identified, determine the appropriate fix. Check the existing documentation to avoid duplicating what's already there:
+
+- `AGENTS.md` — Top-level project instructions and component map
+- `CLAUDE.md` — Entry point referencing AGENTS.md
+- `docs/AGENTS.md` — Documentation site guide
+- `infrahub_sdk/ctl/AGENTS.md` — CLI development guide
+- `infrahub_sdk/pytest_plugin/AGENTS.md` — Pytest plugin guide
+- `tests/AGENTS.md` — Testing guide
+
+Read the relevant existing files to understand what's already documented before proposing changes.
+
+## Step 3: Generate Report
+
+Present the feedback as a structured report with the following sections. Only include sections that have content — skip empty sections.
+
+### Format
+
+```markdown
+## Session Feedback Report
+
+### What I Was Working On
+<!-- Brief summary of the task(s) performed in this conversation -->
+
+### Documentation Gaps
+<!-- Things that should be documented but aren't -->
+
+For each gap:
+
+- **Topic**: What's missing
+- **Where**: Which file should contain this (existing file to update, or new file to create)
+- **Why**: How this would have helped during this conversation
+- **Suggested content**: A draft of what should be added (be specific and actionable)
+
+### Documentation Corrections
+<!-- Things that are documented but wrong or misleading -->
+
+For each correction:
+
+- **File**: Path to the file
+- **Issue**: What's wrong or misleading
+- **Fix**: What it should say instead
+
+### Discovered Patterns
+<!-- Conventions or patterns found in the code that aren't documented -->
+
+For each pattern:
+
+- **Pattern**: Description of the convention
+- **Evidence**: Where in the code this pattern is used (file paths)
+- **Where to document**: Which AGENTS.md or guide file should capture this
+
+### Memory Updates
+<!-- Propose additions/changes to MEMORY.md for cross-session persistence -->
+
+For each update:
+
+- **Action**: Add / Update / Remove
+- **Content**: What to write
+- **Reason**: Why this is worth remembering across sessions
+```
+
+## Step 4: Apply Changes
+
+After presenting the report, ask the user which changes they want to apply. Present the options:
+
+1. **Apply all** — Create/update all proposed documentation files and memory
+2. **Cherry-pick** — Let the user select which changes to apply
+3. **None** — Just keep the report as reference, don't modify any files
+
+
+For approved changes:
+
+- Edit existing files when updating documentation
+- Create new files only when no appropriate existing file exists
+- Update `MEMORY.md` with approved memory changes
+- Keep all changes minimal and focused — don't over-document

dev/commands/pre-ci.md

@@ -0,0 +1,36 @@
+Run a subset of fast CI checks locally. These are lightweight validations that catch common issues before pushing. Run all steps and report a summary at the end.
+
+## Steps
+
+1. **Format** Python code:
+   ```bash
+   uv run invoke format
+   ```
+
+2. **Lint** (YAML, Ruff, ty, mypy, markdownlint, vale):
+   ```bash
+   uv run invoke lint
+   ```
+
+3. **Python unit tests**:
+   ```bash
+   uv run pytest tests/unit/
+   ```
+
+4. **Docs unit tests** (vitest):
+   ```bash
+   (cd docs && npx --no-install vitest run)
+   ```
+
+5. **Validate generated documentation** (regenerate and check for drift):
+   ```bash
+   uv run invoke docs-validate
+   ```
+
+## Instructions
+
+- Run each step in order using the Bash tool.
+- If a step fails, continue with the remaining steps.
+- At the end, print a summary table of all steps with pass/fail status.
+- Do NOT commit or push anything.
+