The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extracts shell commands from .tape files and tutorial manifests and executes them via the shell. Evidence: Step 1.5.4 in SKILL.md uses bash -c to run commands extracted from Type directives in tape files. Evidence: modules/manifest-parsing.md describes a requires field in Playwright components that executes background commands.
[COMMAND_EXECUTION]: The skill performs binary rebuilds using system build tools. Evidence: Phase 1.6.3 in SKILL.md executes cargo install --path . or make build depending on the detected build system.
[EXTERNAL_DOWNLOADS]: The skill documentation suggests installing external dependencies from well-known sources. Evidence: SKILL.md error handling section suggests go install github.com/charmbracelet/vhs@latest and npm install -D @playwright/test.
[COMMAND_EXECUTION]: The validation logic (Phase 1.5.2) executes base commands with the --help flag to verify parameters. Evidence: modules/tape-validation.md Phase 3 extracts commands and runs $base_cmd --help to discover valid flags.
[COMMAND_EXECUTION]: The skill is susceptible to indirect prompt injection via malicious command strings in tutorial assets.
Ingestion points: .tape files and .manifest.yaml files.
Boundary markers: Absent; commands are extracted and executed directly without delimiters or safety instructions.
Capability inventory: Bash tool, timeout 5s bash -c, cargo install, make build.
Sanitization: Absent; the skill relies on sed and grep for extraction without verifying command safety.

tutorial-updates