The Agent Skills Directory

[SAFE]: The skill uses official vendor endpoints (api.scoutos.com) and documentation links (docs.scoutos.com). These are documented as primary functional components.
[EXTERNAL_DOWNLOADS]: Skill instructions include the installation of scoutos packages from PyPI and NPM. These are official SDKs from the verified vendor 'scoutos'.
[COMMAND_EXECUTION]: Includes shell scripts test.sh and test-spawn.sh designed to run the openclaw CLI for testing the skill's logic in local or isolated sessions.
[PROMPT_INJECTION]: Identified a surface for indirect prompt injection within the documented Scout Workflow execution API.
Ingestion points: External data enters the workflow via the inputs field in the POST /v2/workflows/{workflow_id}/execute endpoint described in SKILL.md.
Boundary markers: The provided documentation for Jinja templating does not include specific boundary markers or instructions to disregard embedded instructions.
Capability inventory: The described workflows can execute remote language models, perform HTTP requests, run JavaScript, and interact with third-party platforms like Slack and Discord.
Sanitization: There is no evidence of input validation or sanitization procedures in the documented examples.

scout-workflow