Security audit

Git Release Notes

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for release-note generation, but its helper script can execute local commands if given a maliciously crafted git reference.

Review before installing. Use only on trusted repositories and trusted ref names unless the helper is patched to pass refs to Python as data, such as via environment variables or argv, rather than interpolating them into executable source.

A.I.G

Vulnerability Patterns

Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
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

Findings (1)

T09 · Insecure Skill Coding Practices

Error

Location: scripts/gather_commits.sh:59
Finding: Arbitrary Code Execution Through Git Reference Injection
Content: View full analysis
Remediation: View remediation
5 else "" }) result = { "from_ref": os.environ["FROM_REF"], "to_ref": os.environ["TO_REF"], "commit_count": int(os.environ["COMMIT_COUNT"]), "author_count": int(os.environ["AUTHOR_COUNT"]), "commits": commits } print(json.dumps(result, indent=2)) ' ``` Additional hardening measures: 1. Resolve each input to a canonical object ID with `git rev-parse --verify` and use the resolved IDs for revision traversal where practical. 2. Insert `--` where supported to distinguish revisions from options. 3. Consider enforcing an explicit allowlist for accepted reference syntax if human-readable names must be preserved. 4. Add regression tests using valid Git references containing single quotes, plus signs, parentheses, semicolons, and separator-like text. 5. Verify that unusual commit content cannot corrupt field parsing; preferably use NUL-delimited Git output rather than textual separator strings. ]]>

SkillSpector

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 92% confidence
Finding: The code is related to the general release-notes workflow because it gathers git history between refs, but its actual function is only a low-level data collection step. It outputs structured commit metadata and summary counts, without any analysis, categorization, or formatting into polished release notes or changelogs as described. That is a materially narrower and different behavior than the declared primary purpose, so this is a description-behavior mismatch.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.