The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process OpenAPI specifications provided by users, which are untrusted external data sources. A maliciously crafted specification could contain instructions in metadata or descriptions aimed at influencing the agent's actions during the audit or enrichment steps.
Ingestion points: The agent is instructed to use the Read tool to audit and enrich API specifications in both SKILL.md (Step 1) and references/implementation.md (Step 1).
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores or sanitizes instructions embedded within the specification data.
Capability inventory: The skill has access to Write, Edit, Glob, and a restricted Bash(api:docs-*) tool, providing a functional surface for executing actions requested by the prompt.
Sanitization: No sanitization, structural validation, or safety filtering of the input specification content is specified before the agent processes it.
[COMMAND_EXECUTION]: Reference material in references/examples.md contains examples for automating documentation builds using npx and execSync. While these examples target well-known documentation generators, the use of shell execution for automated tasks is a capability that requires user oversight.
[DYNAMIC_EXECUTION]: The skill provides a reference JavaScript utility (generate-examples.js in references/examples.md) that dynamically constructs shell commands (e.g., curl) and code snippets from API schema data. The script uses string concatenation to wrap JSON-stringified payloads in single quotes, which is a pattern that can be vulnerable to command injection if the input data contains unescaped single quotes and is executed without robust shell escaping.

generating-api-docs