The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing the ecc-agentshield package from npm and references the project repository at github.com/affaan-m/agentshield.
[REMOTE_CODE_EXECUTION]: Recommends running npx ecc-agentshield, which downloads and executes remote code from the npm registry at runtime.
[COMMAND_EXECUTION]: Provides various shell commands for scanning projects, fixing security issues, and initializing new configuration files.
[CREDENTIALS_UNSAFE]: Instructs the user to manually export their ANTHROPIC_API_KEY to the environment for use by the tool's deep analysis feature.
[PROMPT_INJECTION]: The skill analyzes external, potentially untrusted project data, creating a surface for indirect prompt injection.
Ingestion points: Files in the .claude/ directory, including CLAUDE.md, settings.json, and hooks/ scripts (SKILL.md).
Boundary markers: No explicit boundary markers or ignore instructions for the agent are specified in the provided text.
Capability inventory: The skill uses shell command execution (npx) and broad file reading capabilities.
Sanitization: No validation or sanitization mechanisms for the processed configuration content are described.

security-scan