The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references external resources on GitHub (github.com/csark0812/toolbox) to retrieve output schemas and planning verification guidelines. These are maintained within the author's namespace and are used for documentation and formatting consistency.
[COMMAND_EXECUTION]: The skill executes git log and gh pr view commands to perform 'archaeology' on the repository's history and pull request metadata. This data is used by the coordinator to identify recurring 'hotspots' and prevent redundant review cycles.
[PROMPT_INJECTION]: As the skill is designed to process and review user-provided code and plans, it possesses an inherent surface for indirect prompt injection attacks where malicious instructions could be embedded in the data.
Ingestion points: User-specified intent and repository data are gathered into a frozen 'envelope' for review, as described in references/slice-envelope.md.
Boundary markers: The prompt templates use standard markdown headers but lack explicit instructions to the agent to disregard control sequences or specific unique delimiters to isolate the untrusted 'slice' materials.
Capability inventory: The skill can spawn subagents to perform reviews and has access to local CLI tools (git, gh) to gather repository information.
Sanitization: There is no mention of filtering, escaping, or validating the content of the ingested 'slices' before they are included in prompts for the blind review subagents.

iterate