The Agent Skills Directory

[CREDENTIALS_UNSAFE]: Several reference files and documentation templates include hardcoded database connection strings containing literal credentials. While these are likely intended as examples, the use of values such as 'admin:pass' and 'user:pass' rather than secure placeholders constitutes a security risk.
Evidence found in references/reference/06-deployment-runbooks.md: DATABASE_URL=postgres://admin:pass@replica.db.com:5432/tasks.
Evidence found in references/reference/06-deployment-runbooks.md: DATABASE_URL=postgres://admin:pass@primary.db.com:5432/tasks.
Evidence found in references/reference/06-deployment-documentation.md: DATABASE_URL=postgres://user:pass@db:5432/tasks.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, potentially untrusted source code and configuration files to generate documentation. It has significant capabilities including Write, Edit, and Bash access, but lacks proper defensive measures for data ingestion.
Ingestion points: Untrusted data enters the agent context via Read, Glob, and Grep tools when analyzing codebases as defined in references/workflows/brownfield-workflow.md and various mapping files in references/mappings/.
Boundary markers: Absent. The skill instructions do not require the use of delimiters or 'ignore' instructions when processing external content.
Capability inventory: The skill has access to Bash (shell execution), Write, and Edit capabilities across its core documentation generation workflows.
Sanitization: Absent. There is no logic or instruction to sanitize or validate content extracted from external codebases before it is interpolated into prompts or written to the file system.

documentation-specialist