The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides and references scripts for data validation and telemetry. Specifically, README.md and SKILL.md instruct the agent to use scripts/validate_library.py to verify the integrity of content snapshots. Additionally, references/repository-integration.md includes shell commands to run telemetry/version_check.py and telemetry/telemetry_log.py when integrated into a larger repository environment. These scripts are utility-focused and follow standard development practices.
[EXTERNAL_DOWNLOADS]: The documentation in references/production-and-delivery.md includes several links to the Overlap AI documentation (docs.overlap.ai) for API reference and clip prompting. These are informational links to a well-known automation service used to assist the agent in drafting accurate technical briefs, not for remote code execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data sources such as 'content spikes', analytics screenshots, and social media metrics (SKILL.md, references/research-and-learning.md). While this constitutes a potential attack surface for indirect prompt injection, the risk is mitigated by:
- Ingestion points: Data enters through user-provided JSON snapshots and research media.
- Boundary markers: SKILL.md contains explicit instructions to 'label editorial scores as judgments' and to 'distinguish topic signals from post outliers', preventing the agent from blindly following embedded instructions in the data.
- Capability inventory: The skill's primary capabilities are limited to generating markdown tables, production briefs, and running the local validator script.
- Sanitization: The scripts/validate_library.py utility enforces a strict schema for the data interchange format, checking for valid URL schemes, timestamps, and numeric types.

shortform-format-library