The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts untrusted user metrics, such as revenue figures or user counts, which are interpolated into terminal commands used to run the scripts/growth-compass.py calculation script.
Ingestion points: Metric values and data series provided by the user for the --current-value, --previous-value, and --series flags as described in SKILL.md and demonstrated in evals/evals.json.
Boundary markers: The instructions do not provide the agent with specific delimiters or sanitization steps to isolate user-supplied inputs before they are executed via the terminal toolset.
Capability inventory: The skill relies on the terminal toolset. The scripts/growth-compass.py script itself uses only standard Python libraries (math, json, sys, argparse) and performs only arithmetic and output formatting without dangerous system-level access.
Sanitization: The script performs basic input validation by casting metric values to floats using Python's standard library and argparse for flag parsing.

yc-weekly-growth-compass