The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a cost estimation script that reads text content from files in the src/prompts directory, which represents an ingestion surface for untrusted data if prompts are sourced externally.
Ingestion points: scripts/estimate-costs.ts reads the content of all files within the src/prompts directory.
Boundary markers: No specific boundary markers or instructions to ignore embedded content are present in the script.
Capability inventory: The skill utilizes Read, Write, and Bash capabilities.
Sanitization: The script performs a simple length-based calculation (content.length / 4) and does not execute the content, minimizing risk.
[EXTERNAL_DOWNLOADS]: The provided GitHub Action workflows reference official actions from the trusted actions organization.
Evidence: uses: actions/checkout@v4 and uses: actions/setup-node@v4 in the generated workflow files.
[COMMAND_EXECUTION]: The skill utilizes Bash for running tests, validation scripts, and API smoke tests.
Evidence: Use of npx vitest run, npx tsx scripts/estimate-costs.ts, and shell pipelines involving grep, sort, and curl.

mistral-ci-integration