agentleFS
canyonlabzcanyonlabz/perfpilot-hub/.cursor/rules/technology-reference-policy.mdc

perfpilot-hub / rules

Cursor rule5 starsChanged 6 months agoMITView on GitHub

PerfPilot version-aware technical documentation and protocol verification

We split this technology-reference-policy.mdc into the rules it gives an agent

20 rules. About 0 are common, 19 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈19

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before non-trivial technology-specific implementation:

    line 10
  2. Distinctive:

    Read relevant repository documentation, dependency declarations, lockfiles, Dockerfiles, runtime configuration, and existing tests.

    line 12
  3. Distinctive:

    Do not assume a /latest/ page describes the project's approved release.

    line 36
  4. Distinctive:

    If a source is unavailable or a key version is unknown, disclose the gap and ask when it blocks a grounded change.

    line 37
  5. Distinctive:

    Use configured research tools under the project's disclosure/security rules.

    line 40
  6. Distinctive:

    Do not duplicate its detailed version table in this Cursor Rule.

    line 46
  7. Distinctive:

    If the public baseline conflicts with current branch evidence, report the discrepancy and seek guidance rather than silently updating …

    line 50
  8. Distinctive:

    Verify affected message schemas, method names, errors, task states, streaming, discovery, and authentication requirements against the …

    line 64
  9. Distinctive:

    Do not assume an A2A SDK or SDK classes exist, or add an SDK without approval.

    line 66
  10. Distinctive:

    Do not infer A2A compliance from AG2 or AG-UI versions, requirements comments, or a successful smoke test alone.

    line 67
  11. Distinctive:

    Follow the installed AG2 release and existing orchestrator/specialist design.

    line 72
  12. Distinctive:

    Treat AG2 and ag-ui-protocol as distinct packages with distinct version evidence.

    line 73
  13. Distinctive:

    Verify relevant agent lifecycle, tool execution, event ordering, streaming, cancellation, and error behavior against version-appropriate …

    line 74
  14. Very specific:

    Verify the component's package.json and lockfile before selecting APIs.

    line 81
  15. Distinctive:

    Do not import Python logging code into frontend code.

    line 87
  16. Distinctive:

    Keep container secrets out of source, build arguments, image layers, and logs.

    line 105
  17. Distinctive:

    Do not claim an image builds or a container is healthy without actual checks.

    line 106
  18. Distinctive:

    Before running build/start commands or tests, check for implicit migrations, database writes, seed/reset operations, or volume …

    line 107
  19. Distinctive:

    Do not build, start, stop, or modify shared services without appropriate approval.

    line 110
  20. Distinctive:

    Keep useful source URLs, exact releases consulted, and unresolved questions in approved plans or shared engineering documentation as …

    line 114

≈ “Common” is an estimate: at least 1 in 5 of the 6,350 Cursor rule files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
technology-reference-policy.mdc
---
description: PerfPilot version-aware technical documentation and protocol verification
alwaysApply: true
---

# PerfPilot Technology Reference Policy

## Required reference workflow

Before non-trivial technology-specific implementation:
1. Identify the affected component and inspect its existing implementation.
2. Read relevant repository documentation, dependency declarations, lockfiles,
   Dockerfiles, runtime configuration, and existing tests.
3. Consult `docs/engineering/technology-stack.md` (the maintained version and
   official-source inventory) and the relevant sections of
   `docs/engineering/development-workflow.md`. If either is missing from the
   working tree, report the discrepancy; do not invent its contents.
4. Consult authoritative documentation relevant to the proposed change, matched
   to the approved protocol and installed/resolved package versions.
5. Record consequential URLs, versions, assumptions, and verification in the plan.

For a documentation-only change, verify the claims being edited against relevant
implementation and official sources; do not research unrelated technologies.

## Evidence and conflicts

- Approved project policy defines intended architecture and supported versions.
- Dependency constraints define allowed ranges, not exact installed versions.
- Lockfiles, environment/package output, and image configuration provide separate
  evidence about resolved packages and runtimes. State what was actually checked.
- Report discrepancies before implementation; do not silently change dependencies,
  requirements, lockfiles, runtime versions, or public contracts.
- Normative protocol specifications govern wire behavior. Version-matched library
  documentation governs library APIs. Framework examples and AI/search summaries
  do not override the normative specification.
- Do not assume a /latest/ page describes the project's approved release.
- If a source is unavailable or a key version is unknown, disclose the gap and
  ask when it blocks a grounded change. Do not claim to have consulted a URL
  merely because it appears in this file.
- Use configured research tools under the project's disclosure/security rules.

## Technology baseline and verification

- `docs/engineering/technology-stack.md` is the maintained inventory of
  observed versions, declared constraints, components, and official references.
  Do not duplicate its detailed version table in this Cursor Rule.
- An observed local version is not proof of the container version or another
  contributor's environment; inspect the affected manifests, lockfiles, runtime
  output, and image definitions as appropriate and report what was verified.
- If the public baseline conflicts with current branch evidence, report the
  discrepancy and seek guidance rather than silently updating dependencies,
  lockfiles, runtimes, or documentation.
- Preserve AG2 orchestrator/specialist, AG-UI-to-web-UI, external A2A, and MCP
  component boundaries. Do not assume an A2A SDK is installed.

## A2A

- Approved protocol target: A2A v1.0.0. Confirm the binding needed
  for the affected work rather than inferring support from route names.
- Official specification entry point:
  https://a2a-protocol.org/latest/specification/
- Identify a release-specific reference where available and record the release
  consulted. Do not silently follow a future change to /latest/.
- Verify affected message schemas, method names, errors, task states, streaming,
  discovery, and authentication requirements against the applicable specification.
- Do not assume an A2A SDK or SDK classes exist, or add an SDK without approval.
- Do not infer A2A compliance from AG2 or AG-UI versions, requirements comments,
  or a successful smoke test alone. Add/propose focused contract coverage.

## AG2 and AG-UI

- Follow the installed AG2 release and existing orchestrator/specialist design.
- Treat AG2 and ag-ui-protocol as distinct packages with distinct version evidence.
- Verify relevant agent lifecycle, tool execution, event ordering, streaming,
  cancellation, and error behavior against version-appropriate official sources.
- Preserve the backend/frontend integration contract. Do not invent events or
  mix A2A and AG-UI protocol concepts.

## CopilotKit and React

- Verify the component's package.json and lockfile before selecting APIs.
- Match CopilotKit/React integration guidance to resolved frontend packages and
  the backend's supported contract. Do not infer package APIs from Node version.
- The documented frontend baseline has no declared test script or explicit
  test runner. Inspect current tooling before proposing a minimal setup; do not
  install one or assume all frontend verification can be skipped.
- Do not import Python logging code into frontend code. Propose frontend logging
  separately if needed, preserving sensitive-data restrictions.

## FastMCP

- Target the approved version-3 implementation and >=3.4.2,<4 dependency range.
- Consult version-3 documentation or matching source/release tags; do not use
  current-major examples as proof that an API exists in the installed version.
- Preserve established gateway, service, namespace, transport, and tool contracts.
- Future-ready means isolating integration boundaries and recording verified
  migration risks. It does not authorize v4 APIs, speculative compatibility
  shims, dependency upgrades, or a migration.

## Docker and deployment

- Inspect the affected Dockerfile, Compose configuration, environment references,
  build context, startup command, health checks, and documented ports.
- Preserve approved per-component boundaries and declared runtime versions.
- Keep container secrets out of source, build arguments, image layers, and logs.
- Do not claim an image builds or a container is healthy without actual checks.
- Before running build/start commands or tests, check for implicit migrations,
  database writes, seed/reset operations, or volume modifications. The global
  database safety policy applies even when changes occur indirectly.
- Do not build, start, stop, or modify shared services without appropriate approval.

## Research traceability

- Keep useful source URLs, exact releases consulted, and unresolved questions in
  approved plans or shared engineering documentation as appropriate.
- Source comments/docstrings must remain self-contained and free of internal
  work-item or private planning references.
- Rules guide the agent; do not describe their presence as proof of CI enforcement
  or protocol compliance.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.