The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses tls_audit.py to fetch information from crt.sh, which is a well-known and trusted service for searching Certificate Transparency logs. This operation is standard for passive reconnaissance during cryptographic audits.
[COMMAND_EXECUTION]: Several scripts in the skill execute network-related commands, such as tls_audit.py performing TLS/SSH handshake probing and padding_oracle.py making repeated HTTP requests to a target URL. These behaviors are essential and intended for the skill's stated purpose of cryptographic analysis.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating an attack surface for indirect prompt injection if the host agent interprets tool outputs as instructions.
Ingestion points: Multiple scripts (tls_audit.py, rsa_attack.py, ecdsa_lattice.py, padding_oracle.py, jwt_forge.py) ingest and process user-supplied artifacts like public keys, signatures, ciphertexts, and network responses.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potential commands embedded within the cryptographic data or tool results.
Capability inventory: The skill environment provides access to network operations (requests, socket, urllib), file system reads, and intensive mathematical processing.
Sanitization: While scripts validate data formats (e.g., checking PEM structure or JSON schema), they do not sanitize content to mitigate prompt injection risks against the analyzing agent.
[SAFE]: The skill follows security best practices for its domain, correctly identifying itself as an offensive security tool. No malicious code, obfuscation, unauthorized data exfiltration, or persistence mechanisms were detected. The use of external dependencies and services is restricted to well-known and reputable sources.

crypto-analysis