The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements an "Automatic learning" loop that journalizes every invocation to synthesize "playbooks." These playbooks are sequences of shell commands with variable substitution templates ({slot}) that the agent is instructed to execute dynamically based on previous session behavior.
[DATA_EXFILTRATION]: The tool includes a delivery feature (--deliver webhook:<url>) that allows the agent to POST the output of any command—including network health, diagnostic reports, and connected device lists—to an external URL.
[CREDENTIALS_UNSAFE]: The skill provides instructions for harvesting Eero session cookies directly from the user's Chrome browser using the auth login --chrome command, which then stores these credentials locally in a credentials.toml file.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the eero-pp-cli binary from the vendor's official GitHub repository or NPM registry using go install or npx commands.
[PROMPT_INJECTION]: The skill processes untrusted data from the Eero account API which is then stored and retrieved via a local learning loop, creating a surface for indirect injection.
Ingestion points: External network data enters the agent context through the network, device list, and recall commands.
Boundary markers: The skill uses structured JSON output via the --agent flag, but lacks explicit delimiters to prevent the agent from interpreting network metadata (e.g., device names) as instructions.
Capability inventory: The skill has the ability to execute shell commands and perform background network requests.
Sanitization: While the skill documentation mentions stripping personal identifiers (PII) during the teach process, it does not specify methods for escaping or validating external API content before it is interpolated into future prompts.

pp-eero