The Agent Skills Directory

EXTERNAL_DOWNLOADS (HIGH): The skill relies on an external tool called skill-seekers. The documentation points to github.com/jmagly/Skill_Seekers as the source, which is not an approved trusted repository. Executing third-party tools from unverified sources to modify local files presents a significant security risk.
PROMPT_INJECTION (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8). It ingests untrusted data from references/*.md and uses AI to transform this content into documentation that often contains executable code blocks.
Ingestion points: output/<skill-name>/references/*.md (external content processed at runtime).
Boundary markers: Absent. While the skill contains instructional warnings to "not hallucinate," there are no technical delimiters used to separate the instructions from the untrusted data being analyzed.
Capability inventory: The skill performs file system modifications (cp, mv, write) and executes the skill-seekers tool which can further interact with the system.
Sanitization: None detected. The skill directly reads and processes the content of the reference files.
CREDENTIALS_UNSAFE (MEDIUM): The documentation explicitly instructs the user to export ANTHROPIC_API_KEY=sk-ant-.... This practice encourages users to place sensitive API keys in plaintext within their shell environment and command history, increasing the risk of credential exposure.
COMMAND_EXECUTION (LOW): The skill uses shell commands like cp, mv, and grep on paths containing the variable <skill-name>. If the skill name is not properly sanitized, it could lead to path traversal or command injection vulnerabilities.

skill-enhancer