The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to utilize shell-based search utilities to explore and analyze target codebases.
Evidence: The prompt template in references/prompt-template.md instructs the agent to use ripgrep (rg) or grep with the -I flag to efficiently search for security-relevant patterns.
Evidence: The workflow steps in SKILL.md require the agent to generate and save a final threat model report as a Markdown file on the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted source code from external repositories, creating an attack surface for embedded malicious instructions.
Ingestion points: The agent reads local files and extracts architectural information from the user-provided repository path through automated search tools.
Boundary markers: The instructions mandate the redaction of secrets and provide role-play guidance for the agent, but the skill lacks programmatic delimiters to strictly isolate ingested code from its own processing instructions.
Capability inventory: The skill combines file-reading capabilities with the ability to execute shell-based search commands and write report files to the filesystem.
Sanitization: The skill relies on natural language instructions for security hygiene (e.g., "Never output secrets") rather than automated sanitization or filtering of the content being analyzed.

security-threat-model