The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/process.py implements a save_key function that handles the storage of generated private keys. By default, this function uses serialization.NoEncryption() when a passphrase is not explicitly provided. This configuration leads to sensitive CA and end-entity private keys being stored in plaintext PEM format on the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection through its certificate auditing functionality, which ingests and parses untrusted external data.
Ingestion points: The audit_certificate function in scripts/agent.py reads external files provided via the --audit CLI argument (line 62).
Boundary markers: The instructions do not define explicit boundary markers or isolation protocols to prevent the agent from misinterpreting data within the PEM files as instructions.
Capability inventory: The skill has the capability to perform filesystem read/write operations and generate cryptographic materials, which could be misused if the agent is compromised by malicious input data.
Sanitization: While the cryptography library performs structural validation of the X.509 format during parsing, the extracted metadata is passed directly to the agent's context for analysis without secondary filtering.

configuring-certificate-authority-with-openssl