The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads from sensitive file paths and environment variables to perform its audit. Specifically, it parses ~/.claude/settings.json, .env, and ~/.claude.json to extract configuration details. It also executes env | grep to inspect variables such as ANTHROPIC_API_KEY and ANTHROPIC_AUTH_TOKEN, exposing these credentials to the agent context.
[DYNAMIC_EXECUTION]: The skill relies on external scripts bundled within its package (memory-scan.py and cache-report.sh) and uses shell command pipelines to process system information at runtime. This involves executing local Python and Bash scripts to perform file system scanning and log analysis.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes data from potentially untrusted sources within the project directory.
Ingestion points: The skill reads project memory files (CLAUDE.md), agent definitions (~/.claude/agents/*.md), and session logs (~/.claude/projects/*.jsonl).
Boundary markers: Absent. The skill does not use specific delimiters or instructions to ignore embedded prompts in the analyzed files.
Capability inventory: The skill has the ability to execute shell commands and read arbitrary files mentioned in the project chain.
Sanitization: Absent. The skill relies on basic string processing and JSON parsing without specific sanitization for prompt injection.

token-audit