The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple external sources, including the repository source code being reviewed, pull request descriptions, and issue tracker content. This creates an attack surface where malicious instructions embedded in code comments or issue descriptions could attempt to influence the agent's review findings or behavior.
Ingestion points: The skill reads the full content of changed files from the PR head via git worktree, fetches PR metadata via gh pr view, and retrieves spec or issue data via gh issue view or local documentation files.
Boundary markers: The sub-agent prompts provided in the instructions (specifically for the Standards and Spec axes) use structured briefs, but they lack rigid delimiters or explicit 'ignore embedded instructions' directives to isolate the untrusted ingested content.
Capability inventory: The skill has significant capabilities, including the ability to post reviews and comments to GitHub using gh api and the gh pr comment command, as well as performing file system and git operations.
Sanitization: There is no evidence of explicit sanitization or filtering of the ingested content before it is processed by the sub-agents. However, the skill implements a mandatory human-in-the-loop 'approval loop', requiring a user to manually review and approve every finding before any remote action is taken.
[COMMAND_EXECUTION]: The skill executes several shell commands to perform its review tasks, including git worktree, git diff, git log, grep, and various GitHub CLI commands (gh pr view, gh api, gh issue view). These commands are necessary for the tool's core functionality and are executed within the local repository environment.
[EXTERNAL_DOWNLOADS]: The skill fetches remote repository data, code, and pull request metadata from GitHub using git fetch and the authenticated gh CLI. These operations target established GitHub infrastructure and the user's configured remote repositories.
[DYNAMIC_EXECUTION]: The skill employs a conditional logic pattern where it may load and use project-specific agent documentation (docs/agents/issue-tracker.md) to define workflows for fetching issue data. This means the skill's behavior is dynamically influenced by the content of the repository it is analyzing, which is a common but notable pattern in advanced agentic tools.

code-review-multi-axis