The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted input from emails and messaging platforms, which presents a surface for indirect prompt injection. However, this is inherent to the skill's primary purpose of communication triage.
Ingestion points: External messages gathered from email, Slack DMs, LinkedIn DMs, Swan inbox, and CRM task queues in SKILL.md.
Boundary markers: Absent; the instructions do not specify delimiters or instructions to ignore embedded commands within the processed messages.
Capability inventory: The agent can autonomously archive messages, mark CRM tasks as complete, and access internal documentation or codebase logs via other agents as described in the 'Steps' section of SKILL.md.
Sanitization: Absent; no explicit filtering or sanitization of external message content is mentioned before it is processed or used in drafts.
[COMMAND_EXECUTION]: The skill utilizes swan-execute-code to process large snapshots of inbox data when item counts are high. This tool is a vendor-provided resource from the author (swan-gtm) and is used here for efficient data processing rather than arbitrary command execution.

inbox-zero