The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to clone the thinkst/canarytokens-docker repository and interacts with the public canarytokens.org API. These references target the official resources for the deception tool described.
[COMMAND_EXECUTION]: The documented workflow utilizes standard CLI tools including git, docker compose, and curl for deployment, as well as PowerShell commands for creating decoy Active Directory accounts.
[CREDENTIALS_UNSAFE]: The documentation includes a hardcoded password (C0mpl3xDecoy!2026) specifically identified as a decoy credential for Active Directory baiting. This is a legitimate use case for deception technology and does not leak operational secrets.
[DATA_EXFILTRATION]: The scripts/agent.py script records deployment metadata, such as token IDs and URLs, in a local file named canarytoken_inventory.json to allow the user to track decoy assets.
[PROMPT_INJECTION]: The skill maintains an indirect prompt injection surface by ingesting data from the Canarytokens API and storing it locally.
Ingestion points: scripts/agent.py parses JSON responses from the configured Canarytokens server.
Boundary markers: No explicit delimiters or instructions are used to wrap data in the inventory file.
Capability inventory: The script has local file-system write access to maintain the inventory.
Sanitization: The script performs standard JSON parsing but does not sanitize the contents of the API response.

deploying-honeytokens-and-canarytokens