The Agent Skills Directory

[COMMAND_EXECUTION]: The skill allows the execution of arbitrary shell commands through the .portdaddyrc configuration's cmd field and the with-lock tool (e.g., pd with-lock deployment -- npm run deploy).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by design, as it ingests and processes data from other potentially untrusted or compromised agents via the messaging and notes systems.
Ingestion points: add_note, publish_message, get_messages, and check_salvage (as described in SKILL.md and api-reference.md).
Boundary markers: Absent; the skill lacks delimiters or explicit instructions to ignore embedded commands within shared messages or notes.
Capability inventory: Subprocess execution via cmd and with-lock, file system scanning via scan_project, and network operations via tunnels and webhooks.
Sanitization: No evidence of sanitization, escaping, or validation of data retrieved from the inter-agent messaging or note-taking systems.
[DATA_EXFILTRATION]: The webhooks and tunnels features allow the agent to transmit internal data to external URLs or expose local services to the public internet.
[EXTERNAL_DOWNLOADS]: The skill references and depends on the port-daddy NPM package and an external local daemon process.

port-daddy-cli