Corpus
PassAudited by ClawScan on May 1, 2026.
Overview
This skill appears to do what it claims—connect to Corpus to search, read, and add Corpus items—but users should notice that it uses an account token and can write saved links or reminders.
Install only if you want OpenClaw to access your Corpus account. Keep CORPUS_API_TOKEN secret, use a trusted Corpus API base URL, review any retrieved content that may guide code changes, and confirm before the agent saves URLs or creates reminders.
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.
Anyone or any agent process with this token may be able to access supported Corpus data and perform supported Corpus actions.
The skill requires a delegated Corpus API token and uses it for read/write account access. This is expected for the integration, but it is sensitive authority.
Use this skill to read and write a user's Corpus data through the Corpus API. ... `CORPUS_API_TOKEN`: user token for Corpus API access.
Store the token only in the skill environment, do not paste it into prompts or logs, use the Corpus-generated OpenClaw token, and revoke or rotate it if no longer needed.
An unclear request could result in an unwanted saved link or reminder in the user's Corpus account.
The skill exposes account-mutating commands. They are purpose-aligned and documented, with a stated confirmation rule for ambiguous writes.
`save-url --url <url> [--user-note "<note>"]` ... `create-reminder --title "<title>" ...` ... Before write operations (`save-url`, `create-reminder`), confirm user intent if the instruction is ambiguous.
Review the URL, reminder title, description, and scheduled time before allowing write actions, and ask for explicit confirmation when intent is not clear.
Private, stale, or misleading saved content could influence the agent's plan or appear in the conversation if the user asks to use it.
The workflow can bring persistent Corpus content into the agent's context and use it to guide implementation work. This is aligned with the documented use case, but retrieved content should still be treated as reference material rather than automatically trusted instructions.
Use `search` with a focused query. Use `content` for top hits to collect concrete steps. Produce an implementation plan with explicit file changes. Apply code changes in the current working repository after user confirmation.
Limit searches to relevant content, review retrieved items before applying their guidance, and keep the documented confirmation step before repository changes.