The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill downloads a configuration block from the author's GitHub repository (Studio-Moser/skills-n-stuff). This content (AGENTS_Baseline.md) is integrated into local project files via a bundled script (stamp-baseline.sh). While the pattern involves downloading remote content, it serves to synchronize documentation baselines rather than executing arbitrary remote scripts.
[INDIRECT_PROMPT_INJECTION]: The skill scans research reports in specified directories to seed the project backlog and glossary. This behavior establishes a vulnerability surface where untrusted data from these reports could contain malicious instructions designed to influence the agent's project configuration.
Ingestion points: Research/ directories and pulse-config.yaml configuration file.
Boundary markers: None explicitly mentioned for the ingestion of research report content.
Capability inventory: Execution of local scripts, file system write operations, and git remote inspection.
Sanitization: Generated configuration files are validated using a bundled validate-config.sh script.
[COMMAND_EXECUTION]: The skill utilizes several local utility scripts bundled with the plugin (rubric-path.sh, validate-config.sh, fetch-model-data.sh, stamp-baseline.sh) to perform environment checks, validate user configuration, and manage model routing rubrics.
[EXTERNAL_DOWNLOADS]: The skill uses curl to fetch a baseline configuration block and model performance data. These operations target the author's repository and the Artificial Analysis API, respectively.
[DATA_EXPOSURE]: The skill manages model metadata retrieval using an ARTIFICIAL_ANALYSIS_API_KEY, which the user is instructed to configure in their local environment.

setup