The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches code and tools from the author's remote repositories.
Downloads Zig source code from the tkersey/skills-zig repository on GitHub using git clone.
Installs the cas tool using a Homebrew formula from the tkersey/tap tap on macOS.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage its lifecycle and perform its primary functions.
Invokes git clone, zig build, and install to manage the installation of the native CAS tool.
Executes the cas binary and other CLI tools like codex and st with arguments that include user-provided paths and JSON-formatted parameters.
[DYNAMIC_EXECUTION]: The skill performs runtime compilation of source code during its setup phase.
The install_cas_direct function in SKILL.md executes zig build -Doptimize=ReleaseSafe on the downloaded source to generate the executable binary locally.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection via the processing of external data files.
Ingestion points: The skill instructions describe importing data from files like .step/st-plan.jsonl, .step/orchplan.yaml, and .step/mesh-output.csv using the st tool.
Boundary markers: There are no explicit delimiters or instructions to treat the contents of these files as untrusted data.
Capability inventory: The skill possesses capabilities to execute shell commands, compile code, and perform file system operations.
Sanitization: There is no evidence of validation or sanitization of the data ingested from these external files before it is processed.

cas