Add security and moderation agentic workflows

[Evangelink]

Security & Moderation Agentic Workflows

This PR adds two security-focused workflows for automated content moderation and malicious code detection.

Workflows Added

File	Trigger	Purpose
ai-moderator.md	Issue/PR/comment creation	Auto-detects spam, link spam, and AI-generated content
daily-malicious-code-scan.md	Daily schedule	Scans last 3 days of commits for suspicious patterns

AI Moderator

• Triggers on new issues, comments, and PRs from non-maintainer users
• Detects generic spam, link spam, and AI-generated content
• Applies labels (spam, link-spam, ai-generated, ai-inspected)
• Can hide spam comments automatically
• Rate-limited to 5 actions per 60 seconds
• Skips admin, maintainer, write, and triage roles
• Skips known bots (github-actions, copilot, dependabot)

Daily Malicious Code Scan

• Runs daily on schedule and on-demand
• Analyzes files changed in the last 3 days
• Detects: secret exfiltration, out-of-context code, suspicious system operations, obfuscation
• Generates code-scanning alerts visible in the GitHub Security tab
• Threat scoring system (0-10) with severity mapping

Part of the agentic workflows setup series.

[Copilot]

Pull request overview

Adds new gh-aw agentic workflow definitions intended to improve repo hygiene and security posture by automating moderation actions and performing a recurring scan of recent commits for suspicious patterns.

Changes:

• Adds an AI moderation workflow to label/hide suspected spam or AI-generated content on issues/comments/PRs.
• Adds a daily malicious-code scanning workflow definition that can emit code-scanning alerts for suspicious patterns in recent changes.

Show a summary per file

File	Description
.github/workflows/ai-moderator.md	New moderation agent definition (labeling + comment hiding) for issues/comments/PRs.
.github/workflows/daily-malicious-code-scan.md	New scheduled security scan agent definition for analyzing recent commits and emitting code-scanning alerts.

Copilot's findings

Comments suppressed due to low confidence (3)

.github/workflows/ai-moderator.md:37

• The workflow is intended to apply labels and hide comments, but it only requests issues: read / pull-requests: read. Those actions require write permissions (at least issues: write, and pull-requests: write if labeling PRs). Update the permissions block so the safe outputs can be applied successfully.

permissions:
  contents: read
  issues: read
  pull-requests: read

.github/workflows/ai-moderator.md:16

• The instructions mention handling workflow_dispatch (applying labels to the URL specified in the inputs), but workflow_dispatch isn’t listed under on:. Either add a workflow_dispatch trigger (with explicit inputs for the target issue/PR URL/number) or remove the workflow_dispatch-specific instructions to avoid confusion.

on:
  roles: all
  issues:
    types: [opened]
    lock-for-agent: true
  issue_comment:
    types: [created]
    lock-for-agent: true
  pull_request:
    types: [opened]
    forks: "*"
  skip-roles: [admin, maintainer, write, triage]
  skip-bots: [github-actions, copilot, dependabot]

.github/workflows/daily-malicious-code-scan.md:12

• security-events is set to read, but the workflow’s goal is to create code-scanning alerts. Creating/uploading code scanning results requires security-events: write (and typically does not work with read-only permissions). Update the workflow permissions accordingly so alert creation can succeed.

permissions:
  contents: read
  actions: read
  security-events: read

• Files reviewed: 2/2 changed files
• Comments generated: 2

[Copilot]

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 5

[Copilot]

Copilot's findings

Comments suppressed due to low confidence (2)

.github/workflows/daily-malicious-code-scan.md:275

• This section requires each alert to include rule_id, file_path, and start_line, but the create_code_scanning_alert tool in the compiled workflow does not accept those fields. Please align the required fields list with the tool contract (e.g., file, line, message, severity, optional ruleIdSuffix/column) so the agent can successfully report findings.

Your output MUST:

1. **If suspicious patterns are found**:
   - **CALL** the `create_code_scanning_alert` tool for each finding
   - Each alert must include: `rule_id`, `message`, `severity`, `file_path`, `start_line`, `description`
   - Provide detailed descriptions explaining the threat and recommended remediation

.github/workflows/ai-moderator.md:45

• allowed-reasons: [spam] is inconsistent with the Safe Outputs hide_comment tool schema in the generated workflow, which only allows uppercase reasons like SPAM/ABUSE/etc. If the agent follows this config/instruction, safe-output validation may reject the request; please align the allowed reasons and the instruction text with the tool’s accepted values.

    allowed: [spam, ai-generated, link-spam, ai-inspected]
    target: "*"
  hide-comment:
    max: 5
    allowed-reasons: [spam]
  threat-detection: false

• Files reviewed: 11/11 changed files
• Comments generated: 3

[Copilot]

The example create_code_scanning_alert payload uses fields (rule_id, file_path, start_line, description) that don’t match the Safe Outputs tool schema in the generated workflow (expects file, line, message, severity, and optional ruleIdSuffix/column). As written, the agent is likely to emit invalid safe-output calls and fail to create alerts; please update the example and guidance to use the actual parameter names/constraints (notably ruleIdSuffix must match ^[a-zA-Z0-9_-]+$).

This issue also appears on line 270 of the same file.

[Copilot]

The PR description says it adds two workflows (ai-moderator.md and daily-malicious-code-scan.md), but this PR also adds repository-quality-improver.lock.yml (a separate workflow). Please update the PR description to reflect this additional workflow change, or move it to a separate PR if it’s unrelated.

[Copilot]

Comment says "minimum expires: 1 days"; grammar should be singular ("1 day").

Suggested change

	- cron: "37 */2 * * *" # Every 2 hours (based on minimum expires: 1 days)
	- cron: "37 */2 * * *" # Every 2 hours (based on minimum expires: 1 day)