The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources within the codebase, including CONTEXT.md, Architectural Decision Records (ADRs), and files identified via git log analysis.\n
Ingestion points: The Scope and Explore phases read codebase files, documentation (docs/adr/), and domain glossaries (CONTEXT.md).\n
Boundary markers: The instructions do not define explicit delimiters or 'ignore embedded instructions' warnings for the data passed to subagents.\n
Capability inventory: The skill has the capability to write to CONTEXT.md, create new ADR files, and write brief artifacts to the .crank/ directory.\n
Sanitization: There is no evidence of specific sanitization, filtering, or escaping of the ingested content before it is processed by the AI or presented to the user.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute standard version control commands to analyze project history and structure.\n
Evidence: Uses git log --name-only to identify frequently changed files and git rev-parse --show-toplevel to locate the project root.\n
Context: These are routine operations for a developer tool and do not involve untrusted input flowing into the shell commands.

lite-deepen