The Agent Skills Directory

[DATA_EXFILTRATION]: Path traversal vulnerability in scripts/validate_claims.py. The script extracts file paths from untrusted claim text via regular expressions and resolves them against the project root using pathlib.Path. Because pathlib allows absolute paths or directory traversal sequences (e.g., ../../) to override the base path, a malicious claims file can be used to confirm the existence and compute the SHA256 hash of sensitive files on the host system (e.g., /etc/passwd or SSH keys) that the user has permission to read.
[PROMPT_INJECTION]: Indirect prompt injection surface identified through the ingestion of external data.
Ingestion points: The script scripts/validate_claims.py processes claim data from JSON files or standard input which could originate from untrusted sources.
Boundary markers: Absent. There are no explicit delimiters or instructions to the agent to ignore potentially malicious commands embedded in the processed claims.
Capability inventory: The skill includes scripts for file system metadata scanning and file content reading for hashing purposes.
Sanitization: Absent. The script does not validate that paths extracted from claims remain within the intended project directory.

eld-record