The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands specified in verification steps using node:child_process.spawnSync with the shell: true option, allowing for arbitrary command execution.
Evidence: scripts/verify.mjs line 147: spawnSync(step.run, { cwd: root, shell: true, stdio: 'inherit', env });
[DYNAMIC_EXECUTION]: The verification logic dynamically assembles and runs commands from JSON configuration files, including project-level overrides.
Evidence: scripts/verify.mjs merges references/verify.json with .skilly/verify.json and iterates through the resulting steps array to execute the run property of each step.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface where a malicious repository can achieve code execution by providing a crafted .skilly/verify.json file.
Ingestion points: The script scripts/verify.mjs reads .skilly/verify.json and .skilly/config.json from the repository root.
Boundary markers: No delimiters or warnings are present to prevent the execution of malicious instructions embedded in the JSON configuration.
Capability inventory: The skill has full shell execution capabilities via spawnSync in scripts/verify.mjs and file system access.
Sanitization: There is no sanitization or verification of the run command strings before they are passed to the shell.
[EXTERNAL_DOWNLOADS]: Default configurations trigger external downloads of tools and dependencies using package runners.
Evidence: references/verify.json defines steps using npx vitest, npx playwright, and pipx run semgrep.
[REMOTE_CODE_EXECUTION]: The use of npx and pipx allows for the fetching and immediate execution of code from remote registries.
Evidence: Steps in references/verify.json use npx and pipx to run various development and security tools.

verify