The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes common utility commands such as wc, ls, and grep to gather file statistics and verify the presence of project files. These operations are restricted to memory-related directories and the current project path as defined in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes the content of MEMORY.md, which is a file that can be influenced by project content and external data processed by the agent.
Ingestion points: The skill reads file paths from MEMORY.md (located in the Claude projects directory) as described in SKILL.md.
Boundary markers: No specific boundary markers are used when reading data from memory files.
Capability inventory: Uses shell command execution (ls, wc, and the shell file test operator [ -f ]) to check for the existence of files parsed from the memory file.
Sanitization: The shell script uses a restrictive regular expression ([a-zA-Z0-9_/.-]+\.(ts|js|py|md|json|yaml|yml)) to validate file paths before processing. This regex effectively prevents the interpolation of shell metacharacters like semicolons, backticks, or dollar signs that could be used for command injection via the memory file content.

status