The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes build, test, and linting commands (e.g., npm test, pytest, cargo build) defined in the target repository's manifests or scripts during its Phase 0 baseline probing. If a repository being audited contains malicious code in its build or test scripts, the agent would execute it during the audit process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted code and documentation from the repositories it audits to generate its modernization reports.
Ingestion points: Source code files, package manifests (package.json, Cargo.toml, etc.), and documentation files from the target repository.
Boundary markers: The skill uses structured markdown templates (MODERNIZATION_REPORT.md, MODERNIZATION_PLAN.md) and severity-ranked tables to isolate and organize findings.
Capability inventory: Includes shell command execution for baseline verification and dependency scanning, as well as read access to all files within the target repository.
Sanitization: The skill identifies security issues such as hardcoded secrets as part of its audit but does not specify explicit sanitization or filtering of untrusted input before processing by the agent.
[EXTERNAL_DOWNLOADS]: The skill's setup instructions suggest that the user install additional skills from the vendor's public repository (github.com/luongnv89/skills).

codebase-modernizer