The Agent Skills Directory

[COMMAND_EXECUTION]: The utility script scripts/generate_report.py contains a generate_script method that automates the creation of bash scripts (.sh files) using content provided via the --content argument.
[PRIVILEGE_ESCALATION]: Within scripts/generate_report.py, the generate_script method explicitly grants executable permissions to dynamically created files using file_path.chmod(0o755), which can be used to prepare malicious payloads for execution.
[DYNAMIC_EXECUTION]: The skill supports runtime script generation where shell commands are assembled and written to the filesystem, providing a mechanism for executing arbitrary code generated during a session.
[INDIRECT_PROMPT_INJECTION]: The skill instructions in SKILL.md require the agent to scan and process external codebase files, infrastructure configurations, and policy documents, which are untrusted sources.
Ingestion points: Codebase, infrastructure configs, and policy documents in ${CLAUDE_SKILL_DIR}/ are scanned for evidence as per instruction step 3.
Boundary markers: There are no instructions providing delimiters or warnings to ignore embedded instructions within the scanned data.
Capability inventory: The skill has file-writing capabilities and can generate executable scripts via scripts/generate_report.py.
Sanitization: The skill lacks sanitization mechanisms for the data ingested during the scanning process before it is used to generate reports or scripts.

generating-compliance-reports