The Agent Skills Directory

[SAFE]: The skill consists exclusively of Markdown documentation files providing guidance on directory traversal strategies. No executable code, scripts, or binary assets are present.
[INDIRECT_PROMPT_INJECTION]: The described strategy involves the agent reading content from files such as CLAUDE.md, AGENTS.md, and RULES.md encountered during directory traversal. This represents a potential ingestion surface for untrusted data if the agent operates in an untrusted project environment.
Ingestion points: Files matching names like CLAUDE.md, AGENTS.md, START.md, RULES.md, README.md, TODO.md, or custom keys in staircase-config.json (as described in SKILL.md).
Boundary markers: Absent; the documentation does not specify the use of delimiters or 'ignore' instructions for the content of discovered files.
Capability inventory: None; the skill is purely instructional and provides no code or tool definitions.
Sanitization: Absent; no validation or sanitization of the content of the discovered signpost files is mentioned.

staircase-routing