The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses shell evaluation to execute the output of a local binary.
Evidence: eval "$(~/.claude/skills/gstack/bin/gstack-slug 2>/dev/null)" is used to set environment variables or execute commands based on binary output.
Evidence: node -e "..." is used to dynamically parse and execute logic derived from package.json.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses and writes to configuration and history files located in hidden directories within the user's home folder.
Evidence: Extensive read/write operations targeting ~/.gstack/ and ~/.claude/ for tracking health history and configuration state.
Evidence: Use of touch, mkdir -p, and tail on paths in the home directory.
[INDIRECT_PROMPT_INJECTION]: The skill parses untrusted project configuration files to determine which commands to execute, creating a vulnerability surface for indirect injection.
Ingestion points: CLAUDE.md, package.json, tsconfig.json, pyproject.toml, Cargo.toml, and go.mod.
Boundary markers: Absent. No delimiters or warnings are used when processing the content of these files.
Capability inventory: The skill has access to Bash, Write, Edit, and Glob tools, and executes multiple CLI tools including local binaries.
Sanitization: Absent. The skill directly uses values parsed from configuration files (e.g., test scripts from package.json) in shell commands.

health