The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly reads an authentication token from a local configuration file at ~/.config/forgejo/token using the command cat ~/.config/forgejo/token. Storing or accessing plain-text tokens in this manner is a high-risk practice.
[DATA_EXFILTRATION]: The skill uses curl to send the retrieved TOKEN to a remote endpoint. While the URL is not explicitly defined in the provided script snippet (it appears to be missing a target URL in the curl command block), the pattern of reading a secret and immediately passing it to a network tool is a primary indicator of exfiltration risk.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to interact with the local file system and Git, including git commit, git push, and gh issue close. It also pipes network responses directly into Python for execution.
[DYNAMIC_EXECUTION]: The skill uses python3 -c to execute dynamically generated Python code that parses JSON data from stdin. This is used to extract the 'state' and 'statuses' from a remote API response. Executing logic this way bypasses static analysis of a standalone script file.

ship