The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because its core function involves processing and interpreting natural language data from external systems where third parties can provide input.
Ingestion points: The instructions direct the agent to sweep content from source control pull request bodies, issue tracker comments, and real-time chat channels (SKILL.md).
Boundary markers: There are no explicit instructions for using technical delimiters (e.g., XML tags) to isolate untrusted content, though the skill mandates evidence citation and a separation of fact from inference.
Capability inventory: The agent utilizes Model Context Protocol (MCP) tools and shell commands (git log, git blame) to retrieve and process data (SKILL.md).
Sanitization: The skill references a deslop process for final output, but the implementation details and sanitization logic are not defined.
[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to gather investigation metadata.
Evidence: The agent is instructed to run commands such as git log --follow, git blame, and claude mcp list to anchor the investigation and identify available tools (SKILL.md).
[NO_CODE]: The skill consists solely of instructions and configuration metadata. It does not bundle any executable scripts, binaries, or external code dependencies, which limits the risk of direct execution attacks.

inquest