openclaw-secret-scanning-maintainer

Files (2)

• SKILL.md
• secret-scanning.mjs · scripts/

SKILL.md (first lines)

# OpenClaw Secret Scanning Maintainer

**Maintainer-only.** This skill requires repo admin / maintainer permissions to edit or delete other users' comments and resolve secret scanning alerts.

Use this skill when processing alerts from `https://github.com/openclaw/openclaw/security/secret-scanning`.

**Language rule:** All notification comments and replacement comments MUST be written in English.

## Script

All mechanical operations (API calls, temp file management, security enforcements) are handled by:

```
$REPO_ROOT/.agents/skills/openclaw-secret-scanning-maintainer/scripts/secret-scanning.mjs
```

The script enforces:

- `hide_secret=true` on all alert fetches (no plaintext secrets in stdout)
- `mktemp` with random UUIDs for all temp files
- `-F body=@file` for all body uploads (no inline shell quoting)
- Notification templates branched by location type
- Never prints `.secret` or `.body` to stdout

## Overall Flow

Supports single or multiple alerts. For multiple alerts, process in ascending order.

For each alert:

1. **Identify** — `fetch-alert` + `fetch-content` to get metadata and body
2. **Decide** — Agent reads the body file, identifies whether plaintext secrets remain, and produces a redacted version only when needed
3. **Redact** — `redact-body-if-needed` for issue/PR body; skip for comments (delete directly)
4. **Purge** — `delete-comment` + `recreate-comment` for comments; cannot purge body history
5. **Notify** — `notify` posts the right template per location type, unless the current issue/PR body is already redacted
6. **Resolve** — `resolve` closes the alert
7. **Summary** — `summary` prints formatted results

## Step 1: Identify

```bash
# List all open alerts
node secret-scanning.mjs list-open

# Fetch specific alert metadata + locations
node secret-scanning.mjs fetch-alert <NUMBER>

# Fetch content for each location (saves body to temp file)
node secret-scanning.mjs fetch-content '<location-json>'
```

The `fetch-content` output includes:

- `body_file`: path to temp file with full body content
- `author`: who posted it
- `issue_number` / `pr_number`: where it is
- `edit_history_count`: number of existing edits
- `type`: location type for routing
- For `discussion_comment`, it also includes `comment_node_id`, `discussion_node_id`, and `reply_to_node_id` when the original comment was a reply.

### Location type routing

| type                          | Flow                                          |
| ----------------------------- | --------------------------------------------- |
| `issue_comment`               | Comment: delete+recreate                      |
| `pull_request_comment`        | Comment: delete+recreate                      |
| `pull_request_review_comment` | Comment: delete+recreate                      |
| `discussion_comment`          | Discussion comment: delete+recreate (GraphQL) |
| `issue_body`                  | Body: redact in place                         |
| `pull_request_body`           | Body: redact in place                         |
| `commit`                      | Notify only                                   |
| _other_                       | Skip and report                               |

## Step 2: Decide (Agent)

The agent reads the body file from `fetch-content` output and:

1. Identifies ALL secrets in the content (there may be more than the alert flagged)
2. Determines whether any plaintext credential remains in the current body
3. Replaces each remaining secret with `[REDACTED <secret_type>]` — **no partial values, no prefix/suffix**
4. Saves the redacted content to a new temp file

This is the only step that requires semantic understanding. Everything else is mechanical.

For `issue_body` and `pull_request_body`: if the current body has already been redacted by the author and no plaintext credential remains, **do not post a public notification comment**. Resolve the alert with a maintainer-only resolution comment such as:

```bash
node secret-scanning.mjs resolve <ALERT_NUMBER> revoked "Current issue/PR body is already redacted; no public notification posted."
```

This avoids creating a fresh public pointer to historical sensitive content.

## Step 3: Redact

### For comments (issue_comment / PR comments)

**Do NOT redact.** Skip directly to Step 4 (delete + recreate). PATCHing before DELETE creates an unnecessary edit history revision.

### For issue_body / pull_request_body

```bash
node secret-scanning.mjs redact-body-if-needed <issue|pr> <NUMBER> <current-body-file> <redacted-body-file> <result-file>
```

Use the `body_file` from `fetch-content` as `<current-body-file>`. The command writes `notify_required` to `<result-file>` and only PATCHes the body when the redacted file differs from the current body.

## Step 4: Purge Edit History

### Comments — Delete and Recreate

For issue/PR comments:

```bash
# Delete original (all edit history gone)
node secret-scanning.mjs delete-comment <COMMENT_ID>

# Recreate with redacted content
node secret-scanning.mjs recreate-comment <ISSUE_NUMBER> <body-file>
```

From openclaw/openclaw, MIT.

More from openclaw/openclaw