claude-migrate-memory-to-doc
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes several shell commands to automate the migration and verification process.
- It uses
ln -sto create symlinks between tool-specific configuration directories (~/.claudeand~/.codex). - It employs
sed,grep,wc, andfindto read from and modify local configuration files and session logs. - It executes the
codex execcommand to verify that global instruction files are correctly injected into new agent sessions. - [INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow where potentially untrusted data from memory files is promoted to global agent instructions, which could influence future behavior.
- Ingestion points: Data is read from project-specific memory files located in
~/.claude/projects/*/memory/. - Boundary markers: The instructions do not explicitly require structural delimiters or safety markers when inlining memory content into
CLAUDE.md, though a multi-agent review process is recommended. - Capability inventory: The skill possesses capabilities for file writing, symlinking, and shell command execution.
- Sanitization: The skill lacks automated sanitization for migrated text, instead relying on subagent reviews and manual user oversight to identify sensitive data like PII or identity mappings.
Audit Metadata