The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute an internal Python script (scripts/draw_cards.py). This script is used locally to generate the tarot spread. The command parsing in the script is robust, converting the card count to an integer and using strict flag checks, which prevents command injection.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and interpret local markdown files representing tarot card and house meanings. While this is a data ingestion surface, the files are internal to the skill package and contain benign interpretive text. The skill explicitly mitigates risks by instructing the agent that card readings are never sufficient by themselves and must be followed by ordinary engineering evidence in security or correctness contexts.
Ingestion points: JSON output from draw_cards.py and markdown files located in the {baseDir}/cards/ and {baseDir}/houses/ directories.
Boundary markers: The skill uses a structured interpretation workflow where the agent is required to synthesize the 'story' of the spread and document how specific cards influenced their choices.
Capability inventory: The skill is restricted to Bash, Read, Grep, and Glob tools via the YAML frontmatter.
Sanitization: The processed content is static, skill-local documentation. The skill instructions reinforce that card interpretations cannot prove safety, dismiss findings, or replace formal verification.
[SAFE]: The skill demonstrates security best practices, such as using the secrets module for cryptographically secure randomness to avoid bias. It also includes a 'Rationalizations to Reject' section that explicitly prevents the agent from using the skill to bypass safety requirements or user instructions.

let-fate-decide