The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on two bundled shell scripts to perform repository modifications and validations.
scaffold.sh is invoked in the third step to move or copy initial design documents, create the directory structure for living documentation, and configure the check script.
check-guidance.sh is run in the fourth step and by the generated CI workflow to enforce documentation rules using git diff and file existence checks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted design documents (referred to as "frozen artefacts") which could potentially contain malicious instructions.
Ingestion points: scaffold.sh reads the artefact file to extract glossary terms via an awk script. Additionally, the SKILL.md instructions require the agent to read the artefact in full during the work process.
Boundary markers: The instructions do not specify any boundary markers or instructions to ignore embedded commands within the ingested artefacts.
Capability inventory: The skill possesses capabilities to move, copy, and create files, as well as modify shell scripts using sed and perform repository-wide git operations.
Sanitization: The scaffold.sh script uses an awk parser to specifically extract markdown table rows for the glossary, which provides structural filtering but not semantic sanitization of the content.

guidance-tiers