The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script scripts/cap_probe.py to interact with the Abel CAP API. This script uses standard libraries and follows best practices for authentication, including the use of environment variables and scoped configuration files.
[EXTERNAL_DOWNLOADS]: Network operations are directed solely to the official vendor domain cap.abel.ai. This communication is required for the skill's primary function and targets the provider's infrastructure.
[DATA_EXFILTRATION]: Credential management for the ABEL_API_KEY is implemented correctly. The skill does not attempt to access or exfiltrate sensitive files such as SSH keys, cloud provider configurations, or unrelated environment variables.
[PROMPT_INJECTION]: The instructions are focused on guiding the agent through complex causal reasoning steps. There are no patterns suggesting attempts to bypass safety filters or override system-level constraints.
[SAFE]: The skill demonstrates several security-positive patterns, including an automated rendering guard (scripts/render_guard.py) that prevents internal identifiers or raw decimals from appearing in user-facing reports, and a mandatory 'red-team' search step to find evidence contradicting its own conclusions.

abel-ask