The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon setup criteria delegated from other skills, referred to as consumer handoffs.
Ingestion Points: Handoff data provided by calling skills as defined in the references/consumer-handoffs.md schema.
Boundary Markers: Instructions mandate strictly bounded handoffs and utilize explicit status tokens (e.g., Pending, Blocked, Unverifiable) to manage the session state and prevent unintended execution.
Capability Inventory: Capabilities include file system operations (writing/reading restricted .tigerkit/ paths), dynamic script generation, and command execution using npx, bash, and shellcheck.
Sanitization: The skill requires any generated helper scripts to be statically validated with bash -n and shellcheck before being presented to the user.
[EXTERNAL_DOWNLOADS]: The documentation references external tools and repositories for environment configuration.
Package Reference: The skill suggests a conceptual default for browser provider setup using npx -y chrome-devtools-mcp@latest. The use of the @latest tag is an unversioned dependency pattern.
Repository Reference: The references/upstream-distillation.md file acknowledges the skill's origin at https://raw.githubusercontent.com/mattpocock/skills/main/skills/engineering/wizard/SKILL.md.
[DYNAMIC_EXECUTION]: The skill facilitates the creation and verification of runtime-generated scripts.
Evidence: The instructions guide the agent to create one-time "user-run helpers" to assist the user with specific tasks. Safety is managed through mandatory shell syntax checking (bash -n) and static analysis (shellcheck) to ensure the generated code is correct and safe for the user to run.
[COMMAND_EXECUTION]: The skill instructions involve executing shell commands to verify environment state and validate scripts.
Evidence: Utilizes git ls-files and git check-ignore to verify the safety of secret storage paths, and bash -n/shellcheck to validate dynamically created helper scripts.

tk-wizard