The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands extracted from .tape files using bash -c during the pre-flight validation phase to ensure they produce output.
[COMMAND_EXECUTION]: Launches background processes specified in the requires field of .manifest.yaml files during browser session recording.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect injection via project assets. It ingests data from .tape and .manifest.yaml files and uses it as arguments for shell execution without sanitization.
Ingestion points: Manifest files (.manifest.yaml) and VHS tape files (.tape) are parsed in modules/manifest-parsing.md and modules/tape-validation.md.
Boundary markers: Absent; command strings are extracted directly from YAML or tape directives.
Capability inventory: Includes bash -c execution, cargo install/make build commands, and invocation of vhs and playwright tools.
Sanitization: Absent; content is extracted via regex and passed to the shell.
[EXTERNAL_DOWNLOADS]: Fetches and installs the vhs tool from Charmbracelet's GitHub repository using go install and the playwright framework from the official NPM registry.

tutorial-updates