The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill contains instructions for subagents to dynamically locate a script named anysearch_cli.py by probing multiple hidden directories in the user's home folder, such as ~/.claude/skills/anysearch/scripts/, ~/.cursor/skills/anysearch/scripts/, and ~/.anysearch/scripts/. It instructs the agent to locate the script path using shell commands and then execute it using python3 with runtime-generated arguments.
[COMMAND_EXECUTION]: The preamble section executes a shell script check-update.sh using a relative path (../../check-update.sh). This script is located outside the skill's primary directory, leading to the execution of code from an external location.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from web searches and subagent findings without implementing security guardrails.
Ingestion points: Data from WebSearch, mcp__exa__web_search_exa, and parallel subagent results are interpolated into the prompt context.
Boundary markers: No delimiters, markers, or safety instructions are present to mitigate the risk of instructions embedded in the external content.
Capability inventory: The skill possesses extensive capabilities including Bash execution, Write access to the file system, and the ability to spawn further Agent tasks.
Sanitization: The skill does not implement validation, filtering, or escaping for the external data before interpolation.

pm-brainstorm