The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external source code files which are untrusted ingestion points, creating a vulnerability where malicious code content or comments could attempt to influence the agent's behavior.
Ingestion points: Project source files identifying data models and functions (SKILL.md, Step 1).
Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent the agent from being influenced by instructions embedded within the analyzed code.
Capability inventory: The agent can run local commands (ripgrep), execute a shell script (naming-smells.sh), run external linters, write HTML report files, and perform file-system modifications via renaming tools (SKILL.md, Steps 0, 5, and 6).
Sanitization: There is no documented sanitization or filtering of the source code content before it is processed by the agent or integrated into output reports.
[COMMAND_EXECUTION]: The skill executes local commands for static analysis and project auditing.
Evidence: The skill runs ripgrep for pattern matching and utilizes a local bash script scripts/naming-smells.sh for heuristic-based smell detection.
[REMOTE_CODE_EXECUTION]: The skill integrates with external development tools and linters.
Evidence: Instructions include running revive, golangci-lint, eslint, clippy, and ruff. These are well-known and trusted tools within their respective language ecosystems.

naming-review