The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts (scripts/probe_implicit_token_savings.py, scripts/validate.py, scripts/test_skill.py) that utilize subprocess.run to interact with the local environment. These commands are used to identify installed binaries like git, docker, npm, and ripgrep, and to verify that they support the compact output formats recommended by the skill.
[DYNAMIC_EXECUTION]: The script scripts/probe_implicit_token_savings.py dynamically creates a temporary directory structure, including a mock git repository and simple source files, to verify the efficiency of the suggested commands without impacting the user's actual files. Its regression test suite further generates mock shell scripts to simulate tool outputs during testing.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process and summarize output from various shell tools, which represents an ingestion point for external data. While intended for development efficiency, this creates a potential surface for indirect injection if malicious content were present in the command outputs (e.g., a file named with an injection payload). 1. Ingestion points: Markdown instructions in SKILL.md and reference files guide the agent to read and act on the results of commands like ls, rg, and git status. 2. Boundary markers: Absent. 3. Capability inventory: The skill has access to shell execution via the agent's tools, focused on filesystem inspection, git operations, and container management. 4. Sanitization: Not explicitly implemented in the prompt instructions.

implicit-token-savings