The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes source code files through the claudemem tool, which could ingest malicious instructions hidden in code comments or identifiers into the agent's context.
Ingestion points: The agent processes output from claudemem callers, claudemem map, and claudemem test-gaps (defined in SKILL.md), which reflect the contents and structure of the analyzed codebase.
Boundary markers: Absent. There are no delimiters or specific instructions to ignore embedded prompts within the analyzed content.
Capability inventory: The skill allows access to Bash, Task, and Read tools, providing capabilities to execute commands and read local files.
Sanitization: No sanitization or filtering is applied to the data retrieved from the code analysis tool.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform complex diagnostic scripts, including filesystem operations (stat, find, wc) and execution of the claudemem binary with user-defined function names as arguments. It includes logic to calculate stale file counts and verify tool installation status.

tester-detective