The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands (grep, find) upon session start to automatically detect if the current workspace contains a Jetpack Compose project (references/auto-init.md).
[COMMAND_EXECUTION]: During PR reviews, the skill instructs the agent to execute GitHub CLI (gh) and system utility commands (cat, base64) to inspect project settings and codebase conventions (references/pr-review.md).
[EXTERNAL_DOWNLOADS]: The agent is instructed to fetch PR diffs and repository file contents from GitHub using the gh tool when a user provides a PR URL.
[PROMPT_INJECTION]: The skill fetches and processes untrusted data from GitHub repositories without explicit sanitization or boundary markers, creating a surface for indirect prompt injection. A malicious PR could contain code or comments designed to influence the agent's behavior during the review phase (references/pr-review.md).
Ingestion points: PR diffs and repository files fetched via the gh CLI as part of the review workflow in references/pr-review.md.
Boundary markers: Not present; the instructions do not suggest wrapping external content in delimiters or using warnings.
Capability inventory: The agent can execute shell commands (gh, cat, grep, find) and perform multi-file code analysis.
Sanitization: Not present; the agent analyzes the fetched code and metadata directly.

compose-expert