Security risk report for @forgecat/anthropics_claude-plugins-official_claude-md-management v0.0.4
Source Integrity
Low
Profile is sourced from Anthropic's official claude-plugins-official repository on GitHub, a trusted first-party source.
Metadata (author, license, commit hash) are consistent with legitimate open-source publication.
Agent Intent
Low
Content describes legitimate CLAUDE.md maintenance workflows (auditing, updating project context) without injecting malicious instructions or attempting to manipulate the AI agent.
No instructions to read credentials, exfiltrate data, hide instructions, install remote payloads, or leak system prompts.
Guidance on CLAUDE.md quality and update practices is constructive and security-aware (e.g., 'avoid verbose explanations', 'keep it concise', 'actionable commands')—not poisoned or deceptive.
Details
Evidence
The skill explicitly requires user approval before applying changes: 'Ask if the user wants to apply the changes. Only edit files they approve.'
Update guidelines emphasize validation and project-specificity, not weakening security or installing backdoors.
No references to external URLs, remote payloads, or credential exfiltration.
Permissions
PRM-000Medium
The skill declares Bash access, which is a high-capability tool; however its use is narrowly scoped to `find` commands for locating CLAUDE.md files, which is consistent with the stated purpose.
File write (Edit tool) is declared and used, but gated behind explicit user approval before application, reducing effective risk.
The scope of Bash is not fully constrained by the profile itself — a future invocation could theoretically run arbitrary shell commands — making this a medium rather than low concern.
ForgeCat