AGENTS.md · diff
git:20261109.ab63fe5 to git:20261115.5cfb026
1 added, 1 removed. Audit A to A.
What this update changes in the checks
No check changed: every fixed check gives the same answer before and after this update.
A tripped check is a label, not a verdict: read the matched text and the lines below. This week's updates that newly trip a check.
# AGENTS.md — agent-strace development guide
This file tells AI coding agents how to work with the agent-strace repository.
## Project layout
```
src/agent_trace/ Core library — one module per feature
tests/ One test file per module (test_<module>.py)
ADRs/ Architecture Decision Records — read before adding dependencies
docs/ User-facing documentation (setup, commands, production, integrations, security)
examples/ Usage examples for each integration
pyproject.toml Package config and optional extras
```
## Constraints
- **Zero runtime dependencies in `src/agent_trace/`.** Use Python stdlib only. If a feature requires a third-party package, it must be an optional extra in `pyproject.toml` under `[project.optional-dependencies]`, imported lazily inside the function with a clear `ImportError` message. See ADR-0003.
- **Never change the `.agent-traces/` storage format** in a way that breaks existing sessions. Adding new fields is fine. Removing or renaming existing fields requires a major version bump and human approval.
- **Every new CLI command must be registered in `cli.py`** following the existing subparser pattern.
- **Every new feature must have tests** in `tests/test_<module>.py`. Run `python -m unittest discover -s tests -v` to verify.
- **OTLP export uses HTTP/JSON only** — no gRPC. See ADR-0006.
## Development workflow
```bash
# Install in editable mode
pip install -e .
# Run tests
python -m unittest discover -s tests -v
# Run a single test file
python -m unittest discover -s tests -p 'test_watch.py' -v
```
## Adding a new feature
1. Create `src/agent_trace/<feature>.py`
2. Add a `cmd_<feature>` function following the pattern in existing modules
3. Import and register it in `cli.py`
4. Add new `EventType` values to `models.py` if needed
5. Write tests in `tests/test_<feature>.py`
6. Add the command to the appropriate table in `README.md` (one line, linked to docs/)
7. Add full flag reference to `docs/commands.md`
8. If the feature involves production export, update `docs/production.md`
9. If the feature involves a new integration, update `docs/integrations.md`
## Version bumping
- The version lives in `src/agent_trace/__init__.py`. A release must also sync `plugin.json`, `plugins/agent-strace/plugin.json`, `.claude-plugin/plugin.json`, the plugin versions in `.claude-plugin/marketplace.json`, both versions in `.github/plugin/marketplace.json` (`metadata.version` and `plugins[0].version`), and both versions in `server.json` (`version` and `packages[0].version`). Tests fail until they match. `server.json` may lead `__version__` between releases: it names the first PyPI release that ships the current MCP server, and must never name an older one.
+ The version lives in `src/agent_trace/__init__.py`. A release must also sync `plugin.json`, `plugins/agent-strace/plugin.json`, `.claude-plugin/plugin.json`, the plugin versions in `.claude-plugin/marketplace.json`, both versions in `.github/plugin/marketplace.json` (`metadata.version` and `plugins[0].version`), both versions in `server.json` (`version` and `packages[0].version`), and every `Siddhant-K-code/agent-trace@vX.Y.Z` Action pin in `README.md`, `docs/`, and `examples/`. Tests fail until they match. `server.json` may lead `__version__` between releases: it names the first PyPI release that ships the current MCP server, and must never name an older one.
The repo root is a legacy Copilot plugin (`plugin.json` with explicit `agents`, `skills`, `hooks`) and the Claude Code plugin (`.claude-plugin/`, `hooks/hooks.json`). `plugins/agent-strace/` is the Agent Plugins 1.0 copy for marketplaces that require it; its `skills/`, `com.github.copilot/agents/`, and `com.github.copilot/hooks/hooks.json` must match the root `skills/`, `agents/`, and `hooks.json` (tests compare them). Do not add `mcp.json`, `.mcp.json`, or `gemini-extension.json` at the root: see `docs/copilot-plugin.md`.
MCP Registry publishing is automatic. The `mcp-registry` job in `.github/workflows/publish.yml` runs after the PyPI upload succeeds, waits until PyPI serves the version, logs in with `mcp-publisher login github-oidc`, and runs `mcp-publisher publish` from the release tag. It fails if `server.json` does not match the tag version and skips if the Registry already has that version. Manual fallback, from the repo root at the release tag: `mcp-publisher login github` (device auth) then `mcp-publisher publish`. Only publish a version whose PyPI package already exists.
- New feature (new command, new flag, new integration): bump minor (`X.Y.Z` → `X.(Y+1).0`)
- Bug fix or small improvement: bump patch (`X.Y.Z` → `X.Y.(Z+1)`)
- Breaking change to CLI or storage format: bump major — check with maintainer first
## docs/ structure
The `docs/` directory contains user-facing documentation. Keep these files current when adding features:
| File | Contents |
|---|---|
| `docs/setup.md` | Full setup for all 3 integration paths, complete JSON configs |
| `docs/commands.md` | Full flag reference for every command |
| `docs/production.md` | Per-backend OTLP setup (Datadog, Honeycomb, Grafana, etc.) |
| `docs/server.md` | Server-side collector setup, Docker, API reference |
| `docs/integrations.md` | Auto-instrumentation for each framework |
| `docs/mcp.md` | `agent-strace mcp` tools and client setup |
| `docs/vscode.md` | VS Code extension docs (setup, commands, settings) |
| `docs/security.md` | Secret redaction, PII anonymization, policy files |
## README policy
`README.md` is a landing page, not documentation. It must stay under 300 lines. Detailed content goes in `docs/`. Do not add detailed flag descriptions, long examples, or configuration dumps to `README.md`.
## ADRs to read before making architectural decisions
- ADR-0003: Zero runtime dependencies
- ADR-0006: OTLP HTTP/JSON only (no gRPC)
- ADR-0001: Flat event stream data model
- ADR-0002: NDJSON file storage (no database)