OpenClaw Secret Scanning Maintainer

1. Skill Overview

This skill is dedicated to triaging alerts generated by GitHub Secret Scanning. When GitHub detects plaintext credentials (Discord Bot Tokens, Feishu App Secrets, etc.) committed or posted in repository content, it delivers a complete remediation workflow: identify leak locations, redact or delete sensitive material, erase edit history traces, notify affected users, and finally resolve the security alert.

Core Principles:

All notification messages and replacement comments must be written in English only
Maintainer-only permission requirement: Repository admin/maintainer privileges are mandatory to edit/delete third-party comments and resolve alerts
All mechanical operations are script-driven: Every API call, temporary file management, and security enforcement logic is handled by secret-scanning.mjs

2. Core Capabilities

2.1 Alert Discovery & Content Retrieval

The skill first lists all open unresolved alerts, then fetches metadata and exact leak locations for a target alert:

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

# Fetch metadata and leak locations for a specific alert ID
node secret-scanning.mjs fetch-alert <alert_number>

# Pull full content from a single leak location, save raw text to temp file
node secret-scanning.mjs fetch-content '<location_json>'

Output fields from fetch-content:

body_file: Absolute path to temp file holding full raw content
author: Username of the content creator
issue_number / pr_number: Parent Issue or Pull Request identifier
edit_history_count: Existing edit revision count for the target content
type: Location classification used to route subsequent remediation logic

2. Location Type Routing Matrix

Distinct remediation workflows are enforced per content location type:

Location Type	Remediation Logic
`issue_comment`	Delete original comment, rebuild redacted replacement
`pull_request_comment`	Delete original comment, rebuild redacted replacement
`pull_request_review_comment`	Delete original comment, rebuild redacted replacement
`discussion_comment`	Delete + rebuild Discussion thread comment via GraphQL
`issue_body`	In-place redaction via PATCH update
`pull_request_body`	In-place redaction via PATCH update
`commit`	Send user notification only; no cleanup possible
`other`	Skip remediation, log to final summary report

2.2 Analysis & Redaction (Only Semantic Reasoning Step)

The Agent reads the raw temp file output from fetch-content and executes this sequence:

Identify all embedded credentials in the text (not only the single token flagged by the alert)
Verify whether unredacted plaintext secrets still exist in the current live content
Replace every full credential instance with static placeholder [REDACTED] — no partial values, prefixes, or suffixes retained
Write fully sanitized text to a new dedicated temp file

Special rule for issue_body / pull_request_body: If the original author has already manually redacted all credentials with no plaintext remaining, skip public notification comments and resolve the alert silently with an internal maintainer-only note.

2.3 Comment Remediation: Delete + Rebuild (Wipes Entire Edit History)

For all comment-type locations (issue comments, PR thread comments, review comments):

In-place PATCH redaction is forbidden; full delete + rebuild is mandatory
Step 1: Delete the original comment, erasing all historical edit revisions entirely
Step 2: Post a brand-new replacement comment containing the sanitized redacted text

Standard template for rebuilt replacement comments:

Note: The original comment by @{author} has been removed due to secret leakage. Below is the redacted version of the original content.

For Discussion thread comments, GraphQL API endpoints are used for deletion and reinsertion to preserve the original reply thread position.

2.4 Issue / PR Body In-Place Redaction (Edit History Limitation)

For issue_body and pull_request_body:

Run the redact-body-if-needed subcommand to apply inline sanitization via PATCH API

⚠️ Critical Limitation: The edit operation creates a historical revision containing the original plaintext secrets, and there is no API method to purge old edit history versions.
Never publicly advise authors to delete/recreate the Issue or close/reopen the PR — this draws unwanted attention to archived secret-containing history
All guidance about edit history risks is printed exclusively to the maintainer terminal; never included in public-facing comments

Terminal-only maintainer warning output:

⚠️ Issue/PR body edit history still contains plaintext secrets. Contact GitHub Support to purge: https://support.github.com/contact[reference:17]

Hard Rule: Never reference edit history, "edited" buttons, or historical revisions in any public comment or resolution note attached to the alert.

2.5 User Notification Logic

For issue_body / pull_request_body: Send a notification comment only if unredacted secrets were still present and the maintainer performed sanitization. If the author self-redacted all credentials, skip notification and resolve the alert silently.

Automatically selected notification templates mapped to location types:

Comment types: Template referencing "your comment … removed and replaced"
Body types: Template referencing "your issue/PR description … redacted in place"
Commit types: Template referencing "code you committed"

2.6 Resolve Closed Alert

# Default resolution reason: revoked
node secret-scanning.mjs resolve <alert_number>

# Custom resolution note
node secret-scanning.mjs resolve <alert_number> revoked "Custom comment"

Definition of revoked: The credential is deemed compromised, not a confirmation that the user has rotated the secret. The maintainer’s responsibility is limited to removing live plaintext exposure and notifying the affected user; verifying credential rotation falls outside this skill’s scope.

2.7 Final Summary Report

After full remediation completes, export structured JSON results and feed into the summary subcommand:

node secret-scanning.mjs summary /tmp/results.json

The script outputs a bounded result block marked ---BEGIN SUMMARY--- and ---END SUMMARY---. The block must be printed verbatim with no rephrasing, reformatting, abbreviation, or custom rewritten summaries.

3. Primary Use Cases

Triage GitHub Secret Scanning alerts: Rapid end-to-end remediation when plaintext credentials are detected in repository content
Comment secret leak cleanup: Delete secret-bearing comments and rebuild sanitized versions to fully eliminate edit history traces
Issue/PR body secret redaction: Apply inline sanitization and inform maintainers of required GitHub Support contact to purge historical edit revisions
Affected user notification: Automatically send standardized English notifications to users who posted leaked credentials
Batch alert processing: Handle alerts sequentially in ascending order, with optional bulk batch execution on explicit user request
Maintainer privileged operations: All comment deletion, content rebuilding, and alert resolution actions require repository maintainer/admin permissions

4. Mandatory Security Enforcement Rules

Rule	Specification
Separation of duties	Agent only parses text, identifies secrets, and generates redacted content; all mutating API work is delegated to the script
Zero secret exposure	No fragments of any credential appear in public comments, redaction markers, or terminal output logs
Conceal alert metadata	Alert URLs and numeric alert IDs are never printed in public comments
Comment remediation policy	Skip inline PATCH edits for comments; full DELETE + rebuild is mandatory to avoid persistent edit history
Edit history confidentiality	Never mention edit history, "edited" UI controls, or commit SHA values in any public-facing text
Deletion confirmation gate	Explicit confirmation is required before deleting any user comment
Sequential alert handling	Process alerts one at a time unless bulk batch processing is explicitly requested
English-only public text	All automated notification and replacement comments use English exclusively
Unsupported type logging	Skip unsupported location categories and log the omission clearly in the final summary

Summary

OpenClaw Secret Scanning Maintainer is a dedicated maintainer skill for responding to GitHub Secret Scanning security alerts within the OpenClaw repository. It implements a full closed-loop workflow: alert enumeration, raw content retrieval, automated credential redaction, comment deletion & reconstruction, inline Issue/PR body sanitization, standardized user notification, and formal alert resolution. It combines scripted automated API operations with lightweight semantic analysis by the Agent to fully eliminate live plaintext secret exposure, while enforcing strict security guardrails prohibiting disclosure of credential fragments, alert identifiers, and sensitive edit history details. This skill is designed for OpenClaw maintainers to triage and remediate secret leak alerts triggered by GitHub’s automated secret scanning feature.