The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The pr-prepare agent processes untrusted external data that could contain malicious instructions designed to manipulate the agent's output or actions.
Ingestion points: The agent reads the content of CONTRIBUTING.md, .github/CONTRIBUTING.md, CODE_OF_CONDUCT.md, package.json, README.md, and CHANGELOG.md using the cat and grep tools to extract project standards and check for documentation updates.
Boundary markers: The instructions lack explicit boundary markers or directives to treat the content of these files strictly as data, which may allow embedded instructions in those files to override the agent's primary task.
Capability inventory: The skill utilizes the Bash tool to execute project-specific scripts including npm run lint, npm test, npx eslint, npx prettier, and npx tsc. It also suggests git and gh (GitHub CLI) commands to the user.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content read from the external repository files before processing.
[COMMAND_EXECUTION]: The skill provides scripts and instructions that execute shell commands and development tools.
The scripts/pre-pr-check.sh and scripts/clean-branch.sh files use standard git and file system utilities to validate and clean the local repository.
The pr-prepare agent executes common development tools like npm, npx, and git to verify project compliance. These are standard operations for a development-focused skill and target well-known ecosystem tools.

open-source-contributions