The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow example provided in SKILL.md instructs the execution of a local script named smtp_relay_server.py using python3. This script is not provided within the skill's file list, meaning its contents and security posture are unverifiable.
[DATA_EXFILTRATION]: The skill utilizes the pilotctl send-file command, which allows the agent to read and transmit local files. If an attacker can influence the file path argument, this could lead to the exfiltration of sensitive files such as SSH keys or environment configurations.
[PROMPT_INJECTION]: The skill is designed to receive and process external messages via pilotctl inbox and pilotctl recv. This creates a significant surface for Indirect Prompt Injection, where malicious instructions embedded in incoming emails could manipulate the agent's behavior.
Ingestion points: pilotctl inbox, pilotctl recv (documented in SKILL.md)
Boundary markers: None identified in the provided instructions.
Capability inventory: Execution of shell commands via Bash, file system access via send-file, and network configuration via set-webhook.
Sanitization: No evidence of sanitization or content filtering for incoming message data.
[COMMAND_EXECUTION]: The skill requires the Bash tool and executes multiple subprocesses via the pilotctl binary, including starting a daemon and managing webhooks.

pilot-email-bridge