The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core workflow involves reading and analyzing existing codebases using Read, Grep, and Glob tools to ground documentation claims. This creates a surface where malicious instructions embedded in source code (e.g., in comments or string literals) could potentially influence the agent's behavior during the drafting process.
Ingestion points: Reads project files and external sources via Read, Grep, Glob, and WebFetch (SKILL.md, Step 2).
Boundary markers: The skill explicitly instructs to extract "word-for-word quotes" before analysis to ground responses, which serves as a basic delimiter (SKILL.md, Step 2.1).
Capability inventory: The skill has access to Bash, Write, Edit, and WebFetch tools as defined in the frontmatter.
Sanitization: Relies on independent verification passes and instructions to retract unsupported claims.
[COMMAND_EXECUTION]: The instructions direct the agent to use the Bash tool to perform active verification tasks in the environment.
Evidence: Step 3 suggests the agent should "run the test, execute the snippet, curl the endpoint" as an external verification step (SKILL.md).
[DYNAMIC_EXECUTION]: The workflow encourages the agent to execute code snippets found in the codebase or generated during the documentation process to verify accuracy.
Evidence: "A tool result beats any amount of model deliberation. ... execute the snippet" (SKILL.md, Step 3).

tech-writing