The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (project files, knowledge captures, and verified readmes) to generate documentation, creating an attack surface where malicious instructions in those files could influence agent behavior.
Ingestion points: The skill reads project-specific instructions, root READMEs, Git status, and external data types like project-knowledge-capture and verified-readme as specified in the 'Assignment' section of SKILL.md.
Boundary markers: There are no instructions to wrap ingested content in delimiters or to ignore potential instructions embedded within the source data.
Capability inventory: The skill possesses 'local-write' authority to create or update files and the ability to inspect CI, scripts, and package metadata as described in the 'Select the files' and 'Assignment' sections.
Sanitization: The skill lacks explicit instructions for escaping or validating external content before it is interpolated into the orientation documents.
[COMMAND_EXECUTION]: The skill authorizes the execution of local shell commands to verify documentation claims and project state.
Evidence: The 'Write complementary files' and 'Verify and report' sections instruct the agent to "Run a safe narrow command when practical" and ensure that "applicable commands... pass their checks."

kata-orient