DeepRecall
PassAudited by ClawScan on May 10, 2026.
Overview
DeepRecall appears purpose-aligned, but users should know it reads local memory/workspace files and uses local LLM provider credentials to send selected content to configured model APIs.
Install only if you are comfortable with a memory-recall tool reading your OpenClaw workspace and using your configured LLM credentials. Start with narrow scopes, avoid running broad `all` searches on secret-heavy projects, and verify your provider/base URL settings.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The skill may use existing local LLM account credentials, which could incur provider usage or expose account-linked activity to the configured provider.
The provider bridge reads local OpenClaw configuration and cached provider tokens, including a GitHub Copilot token, to authenticate LLM requests.
CONFIG_FILE = OPENCLAW_DIR / "openclaw.json"
AUTH_PROFILES_FILE = OPENCLAW_DIR / "agents" / "main" / "agent" / "auth-profiles.json"
CREDENTIALS_DIR = OPENCLAW_DIR / "credentials"
...
token_file = creds_dir / "github-copilot.token.json"
...
token = data.get("token")Use dedicated or limited-scope API keys where possible, review `~/.openclaw` provider settings, and verify any custom model base URL before use.
Private memory or project content may be transmitted to the configured LLM provider when broad scopes are used.
The skill discloses that it can search broad workspace scopes and send recall context to an external OpenAI-compatible LLM endpoint.
Just markdown files and HTTP calls to any OpenAI-compatible LLM endpoint. ... `project` | All readable workspace files ... `all` | Identity + memory + project (everything)
Prefer narrow scopes such as `identity` or `memory` for sensitive work, avoid `all` on workspaces containing secrets, and review the privacy terms of the configured LLM provider.
Sensitive names or topics from memory files can be summarized into a persistent index, and stale or poisoned memory entries could influence later recall results.
The indexer extracts people, topics, and other metadata from memory files and persists them in `MEMORY_INDEX.md`.
lines.append("## People")
...
lines.append("## Topics")
...
index_path = workspace / "MEMORY_INDEX.md"
index_path.write_text(index_content)Review generated memory indexes periodically, remove untrusted or sensitive files from the workspace, and regenerate the index after cleanup.