The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill contains instructions to install unverified packages from public registries without version pinning, specifically npm i -g padding-oracle-attacker, pip install ciphey, and pip install xortool. These are not from trusted sources and lack integrity verification.
[DATA_EXFILTRATION] (LOW): The factordb function in reference/rsa-attacks.md makes GET requests to http://factordb.com/api. This transmits potentially sensitive cryptographic moduli to a non-whitelisted external service.
[COMMAND_EXECUTION] (LOW): The skill assumes the presence of and uses the Bash tool to execute several local binaries like yafu, sage, and RsaCtfTool.py for its primary cryptographic tasks.
[PROMPT_INJECTION] (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8). Evidence:
Ingestion point: Untrusted ciphertext is passed to tools like ciphey in SKILL.md.
Boundary markers: None present to distinguish tool output from instructions.
Capability inventory: High-risk tools Bash, Read, and Write are enabled.
Sanitization: No sanitization or escaping of processed cryptographic output is performed.

crypto-analysis