The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to use shell commands (Bash) to gather context, including 'git log' for history and 'npm test' for verifying failing tests (references/examples.md).
[EXTERNAL_DOWNLOADS]: The skill uses 'WebSearch' and 'WebFetch' to retrieve information such as documentation and best practices during the research phase (references/research-strategies.md).
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection (Category 8) because it incorporates data from external websites and local files into the agent's decision-making process for generating questions. Ingestion points: user prompts, local files (references/research-strategies.md), and web content (references/research-strategies.md). Capability inventory: shell execution (Bash), file system access (Grep, Glob, Read), network operations (WebFetch, WebSearch), and interactive tool use (AskUserQuestion). Boundary markers: no explicit markers or warnings were identified in the instructions for isolating ingested research content. Sanitization: no specific sanitization or validation logic for the content retrieved during research was identified.

prompt-improver