The Agent Skills Directory

[SAFE]: The skill performs legitimate prioritization calculations using the RICE framework (Reach, Impact, Confidence, Effort). The logic is transparent and focuses solely on assisting users with project management decisions.
[COMMAND_EXECUTION]: The skill uses a bundled Python script scripts/rice_prioritizer.py. Analysis of the script shows it performs safe operations: it reads a local JSON file specified via a command-line argument, performs basic arithmetic to calculate RICE scores and cost-of-delay, and prints a formatted table to the console. There are no calls to subprocess, os.system, or any network-based operations.
[DATA_EXFILTRATION]: No network operations or sensitive file access patterns were detected. The script only reads the configuration file provided to it and does not attempt to access credentials, environment variables, or remote servers.
[PROMPT_INJECTION]: The instructions in SKILL.md are descriptive and guide the agent on how to use the methodology. There are no attempts to override system prompts, bypass safety filters, or hide instructions via obfuscation.
[INDIRECT_PROMPT_INJECTION]: The skill processes initiative data provided in a JSON format. While this is external data, the script treats it as strictly numerical or string data for calculation and display. There is no execution or interpolation of this data into a prompt that could trigger unintended agent behavior. The vulnerability surface is minimal and handled by standard JSON parsing.

initiative-prioritizer