Skip to content
Agent Permission Protocol (APP): Execution-Time Authority for AI Agents
v0.2.0

Contributing to the Agent Permission Protocol

The Agent Permission Protocol (APP) is a public specification stewarded by Crittora.

This repository exists to define, refine, and preserve a minimal set of invariants for execution-time authority in agentic systems. Contributions are welcome in the form of critique, clarification, and proposed changes, provided they advance the protocol’s clarity, correctness, or enforceability.

For governance, canonical publication rules, and change classification, see the repository GOVERNANCE.md.

Scope of contributions

Constructive contributions include:

  • Identifying ambiguities or underspecified behavior
  • Highlighting edge cases or failure modes
  • Proposing clarifications to definitions or invariants
  • Suggesting improvements to protocol structure or terminology
  • Pointing out inconsistencies between sections

Contributions should focus on protocol semantics, not on implementation preferences, tooling choices, or product considerations.

How to contribute

Issues

Feedback and discussion should begin as GitHub Issues.

Use the issue templates when possible:

  • Protocol change for semantic or conformance proposals
  • Clarification request for ambiguity, inconsistency, or unclear wording

When opening an issue:

  • Be specific and precise
  • Reference the relevant section or concept
  • Describe the problem or ambiguity clearly
  • Avoid speculative or purely theoretical critique without grounding

Issues may be used for:

  • Clarification requests
  • Problem identification
  • Design discussion
  • Proposed changes

Pull requests

Pull requests may be submitted for:

  • Text clarifications
  • Corrections
  • Structural improvements
  • Narrowly scoped protocol changes with clear rationale

Before submitting a pull request:

  • Open an issue describing the proposed change
  • Explain the rationale and expected impact
  • State whether the change is normative or non-normative
  • Describe interoperability and security impact
  • Keep changes narrowly scoped

Pull requests may be accepted, modified, or declined at the steward’s discretion.

Editorial authority

Final editorial authority for the Agent Permission Protocol rests with the steward.

This includes decisions regarding:

  • Protocol semantics
  • Inclusion or exclusion of features
  • Versioning and release timing
  • Resolution of disputes or conflicting proposals

This authority exists to preserve protocol coherence, semantic stability, and long-term integrity.

Versioning and change management

The Agent Permission Protocol is versioned.

  • Substantive semantic changes will result in a new version
  • Clarifications that do not alter semantics may be included within a version
  • All changes are documented in the CHANGELOG.md

See the repository GOVERNANCE.md for release discipline and version classification.

Backward compatibility is considered a protocol property and is treated conservatively.

Code of conduct

Participation in this repository is expected to be professional and respectful.

Debate is encouraged.
Dismissiveness, personal attacks, or bad-faith engagement are not.

Implementation discussion

This repository is not intended for:

  • Product promotion
  • Vendor comparison
  • Tooling debates
  • Marketing or evangelism

Implementation discussions may occur only insofar as they illuminate protocol semantics or failure modes.

Closing note

The goal of this repository is not consensus.

It is correctness.

The Agent Permission Protocol is intended to be scrutinized, challenged, and improved through rigorous discussion. Contributions that strengthen the protocol’s invariants are welcome.