The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes numerous shell commands (such as gh, git, and curl) and Node.js/Python scripts. These operations are required to perform file resolution, execute automated audit checks, and interact with the CLI tools of integrated platforms.
[DATA_EXFILTRATION]: The skill reads sensitive information from environment variables and local files (e.g., .env, .secrets/tooling.env) to authenticate with external services like Jira and Bitbucket. This behavior is necessary for its primary function of synchronizing task data, and it includes mechanisms to redact secrets from logs and report artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text from GitHub issue bodies, PR comments, and external web pages. It defines an explicit 'untrusted-content rule' which instructs the agent to treat this content as data rather than instructions, mitigating the risk of adversarial control through prompt injection.
[DYNAMIC_EXECUTION]: The skill utilizes eval in bash scripts for internal variable management and executes an embedded Python program to handle complex YAML configuration parsing when specialized libraries are unavailable.

review-task