The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive local file paths to extract credentials and configuration data during the migration process.
Evidence: The scripts/openclaw_to_hermes.py script reads from ~/.openclaw/.env, ~/.openclaw/openclaw.json, and the ~/.openclaw/credentials/ directory to retrieve API keys and tokens for services including Telegram, OpenAI, and Anthropic.
[PROMPT_INJECTION]: The skill ingests untrusted data from the user's source environment and incorporates it into the agent's persona and memory, presenting an indirect prompt injection surface.
Ingestion points: The scripts/openclaw_to_hermes.py script reads user-supplied content from SOUL.md, MEMORY.md, and USER.md (as described in SKILL.md).
Boundary markers: Absent. There are no specific delimiters or instructions provided to the agent to distinguish between migrated content and system instructions during interpolation.
Capability inventory: The skill has capabilities for file system writes and core configuration modification (.env, config.yaml) within the scripts/openclaw_to_hermes.py script.
Sanitization: Basic string replacement for rebranding is performed via the rebrand_text function, but no robust sanitization or validation of the ingested text is implemented to prevent instruction injection.
[COMMAND_EXECUTION]: The skill requires the agent to execute a local Python script with command-line arguments to perform migration tasks.
Evidence: SKILL.md provides instructions and examples for the agent to use the terminal tool to run python3 scripts/openclaw_to_hermes.py with various execution flags.

openclaw-migration