ClawHub
Back to skill

Security audit

gh-issues

Security checks for vulnerabilities and agentic risk

Overview

This skill is purpose-built for GitHub issue automation, but it can autonomously modify repositories, push branches, open PRs, and process untrusted GitHub content without strong approval or trust-boundary controls.

Install only if you intend to let the agent operate on GitHub repositories with write-capable credentials. Use dry-run first, avoid --yes/--cron/watch on repositories with untrusted issue or review authors, and prefer a minimally scoped GitHub token plus manual review before pushes and PR updates.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:137
Finding

Untrusted GitHub Content Can Hijack Autonomous Repository Workers

Content
View full analysis
`. - required proof and PR body. - notification route. Worker instructions: ```text Use gh and git. Do not handwave. Checkout/create fix/issue- from BASE_BRANCH. Implement minimal fix. Run relevant tests. Commit with conventional message. Push to PUSH_REMOTE. Open PR against SOURCE_REPO BASE_BRANCH. PR body: What Problem This Solves + Why This Change Was Made + User Impact + Evidence + visible Fixes SOURCE_REPO#. Report PR URL or failure reason. Send completion/failure with openclaw message send if route provided. ``` ``` The review workflow contains the same trust-boundary issue: ```markdown Fetch review threads/comments: ```bash gh pr view --repo "$SOURCE_REPO" --json url,headRefName,comments,reviews gh api "repos/$SOURCE_REPO/pulls//comments" gh api "repos/$SOURCE_REPO/issues//comments" ``` Only process `fix/issue-*` PRs created by this workflow unless the user explicitly named PR numbers. Group actionable comments by PR. Ignore praise, status, duplicates, and already-addressed comments. Spawn one worker per selected/scoped PR, same background rules. Review worker instructions: ```text Checkout PR branch. Read all actionable review comments. Patch minimal changes. Run relevant tests. Commit and push normally; do not force-push unless explicitly told. Reply to addressed comments with fix + commit/file reference. Report comments addressed/skipped and proof. ``` ``` ### Technical Analysis The Skill requires issue titles, bodies, labels, and review comments to be placed into autonomous worker context. Th ...[truncated 3102 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill is user-invocable and describes a broad 'issue-to-PR automation' workflow without tight activation boundaries or explicit negative examples. In practice, this can cause the agent to trigger on loosely related requests and proceed into repo inspection, issue enumeration, branch creation, worker spawning, and PR activity that may exceed the user's intended scope.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The skill explicitly supports autonomous execution modes like '--yes', '--cron', '--watch', background worker spawning, PR creation, and review-comment processing with reduced or no interactive confirmation. Because the workflow can make code changes, push branches, and open or update PRs across repeated polling cycles, misuse or accidental invocation could lead to unauthorized repository modifications and sustained automated actions.

Content

Scanner excerpt · SKILL.md (reported line 41)May include surrounding context.

md
- `--watch`: poll issues + reviews.
- `--interval <minutes>`: default 5.
- `--dry-run`: list only.
- `--yes`: no confirmation.
- `--reviews-only`: skip issue fixing; handle PR reviews.
- `--cron`: spawn and exit; implies `--yes`.
- `--model <id>`: pass to workers when supported.

Static analysis

No suspicious patterns detected.