The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) and Git commands to inspect Pull Request metadata, fetch remote branches, manage worktrees, and perform merge or rebase operations. These actions are performed within the scope of the user's authorized repository access.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, potentially untrusted sources such as Pull Request branch content and conflict hunks. This creates a surface for instructions embedded in code or comments to influence the agent's behavior.
Ingestion points: Git conflict hunks, source code from the PR and base branches, and metadata retrieved via gh pr view in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded within the code being resolved.
Capability inventory: The agent has capabilities to execute shell commands (gh, git, build tools), modify files, and push changes to remote branches as detailed in SKILL.md.
Sanitization: No explicit sanitization or filtering of the code content or conflict text is described before the agent processes or resolves it.
[DYNAMIC_EXECUTION]: The skill mandates running the repository's build, format, lint, and test suites to verify conflict resolutions. This involves executing arbitrary scripts defined in the repository (e.g., via npm, cargo, or poetry as mentioned in references/conflicts.md), which could be controlled by a malicious PR author.

resolve-pr-conflicts