Back to skill

Security audit

Justin Ex Clapback Writing Style

Security checks for vulnerabilities and agentic risk

Overview

The skill is a text-only Chinese writing-style helper, but the published package includes private maintainer memory and local development metadata that should not be in an installable artifact.

Review before installing or redistributing. The writing skill itself appears non-executable and purpose-aligned, but the publisher should remove `.workbuddy/`, sanitize local paths and maintainer workflow notes, and rebuild the release from an explicit allowlist.

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

Warning
Location
.workbuddy/memory/2026-08-28.md:1
Finding

Private Development Memory and Maintainer Metadata Included in the Artifact

Content
View full analysis

Vulnerability Details

File Location: .workbuddy/memory/2026-08-28.md, lines 1–23
Vulnerability Type: Sensitive development metadata exposure
Risk Level: Medium

Evidence

markdown
# 2026-08-28

## SkillForge specification audit — ex-clapback-writing-style

- Audit this skill using SkillForge Audit-Only (L3+L4+L5).
...
## Push to GitHub
- Target repository: https://github.com/qomob/ex-clapback-writing-style (org: qomob, gh logged in, protocol https).
- Process: directory was not originally a git repository → `git init -b main` → add remote → fetch discovered remote main already had an initial commit (v1.1.0 original, without frontmatter).
- Committed local fixes as `4f2b98b`, then `git merge -X ours origin/main --allow-unrelated-histories` ... merge commit `fc74299`.
...
- Push successful: `8ed7312..fc74299 main -> main`, remote HEAD = fc74299.

The repository itself states that this directory is private and must not be published:

gitignore
# Local tooling / private memory — never publish
.workbuddy/

An additional public-facing audit artifact exposes a local maintainer path:

markdown
> Audit target: `/Users/jonki/Desktop/skill/ex-clapback-writing-style`

Technical Analysis

The packaged artifact contains an agent/tool memory file that the project explicitly identifies as private. The file discloses internal development history, repository ownership, Git workflow, commit identifiers, tooling state, and the fact that the GitHub CLI was authenticated during development. skillforge-audit.md additionally reveals a local account name through an absolute filesystem path.

No password, access token, private key, session cookie, or other directly reusable credential was found. The statement that GitHub CLI was logged in does not expose its authentication material, and the Skill contains no mechanism that reads or transmits these files at runtime. The issue is therefore an information-disclosure and packaging-hygiene weak ...[truncated 1349 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove .workbuddy/ from the distributed artifact and all release archives.
  2. Purge previously published copies from package releases and repository history where practical.
  3. Replace absolute local paths in public reports with neutral project-relative paths.
  4. Remove Git authentication-state notes, internal commit workflow details, and other nonessential operational metadata from release artifacts.
  5. Build releases from an explicit allowlist, such as:
    • SKILL.md
    • runtime.md
    • Required files under style/, references/, and tests/
    • Sanitized public documentation
  6. Add a CI release gate that rejects:
    • .workbuddy/ and other memory/state directories
    • Absolute home-directory paths
    • Credential-like strings
    • Private keys, tokens, cookies, and environment dumps
    • Editor, agent, and local tooling state
  7. Verify the final archive contents after packaging rather than relying solely on .gitignore, since ignore rules do not automatically exclude files from every packaging process.
  8. Credential rotation is not required based on the audited content alone. Rotate credentials only if a broader review discovers authentication material in prior releases or repository history.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (15)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
65% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · .workbuddy/memory/2026-08-28.md (reported line 20)May include surrounding context.

md
## 推送到 GitHub
- 目标仓库:https://github.com/qomob/ex-clapback-writing-style (org: qomob,gh 已登录,protocol https)。
- 流程:目录原非 git 仓库 → `git init -b main` → 加 remote → fetch 发现远程 main 已有初始提交(v1.1.0 原版,无 frontmatter)。
- 提交本地修复为 `4f2b98b`,再 `git merge -X ours origin/main --allow-unrelated-histories`(冲突文件 SKILL.md/runtime.md 以本地修复版为准),合并提交 `fc74299`。
- 加 `.gitignore` 排除 `.workbuddy/`(本地记忆/隐私)与 `.DS_Store`,确认未被跟踪。
- 已验证合并后 SKILL.md 含 frontmatter + version 1.2.0、runtime.md 含 golden-set 引用、golden-set/audit 已跟踪、`.workbuddy` 未泄露。
- 推送成功:`8ed7312..fc74299 main -> main`,remote HEAD = fc74299。

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The description says the skill is used to write Chinese social-media copy, which imposes a language constraint as part of the skill behavior. Under the policy rule, forcing a specific language without user opt-in or a clearly justified region-specific limitation is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

This line specifies that the generated content must be Chinese social-media copy, reinforcing a mandatory language policy. Because no user language choice or explicit justified locale scope is provided here, it conflicts with the language/locale policy criteria.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This manifest repeatedly specifies Chinese-language naming and a China-specific audience/culture context, including a Chinese skill name, Chinese social platforms, and '中文互联网' culture. Because the file does not offer a language/locale choice or clearly justify the restriction as a region-specific tool, it creates a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file contains natural-language content that assumes a Chinese-language operating context throughout, including the title, evaluation narrative, and stated target user. Under the policy rule for language/locale, forcing or presuming a specific language without offering choice or documenting a justified regional limitation is a violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The runtime instructions explicitly require output in Chinese and prescribe a fixed stylistic format, without any mechanism to respect the user's language preference. This can override higher-level user expectations or platform norms, causing unauthorized language coercion and reducing user control over outputs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The entire audit report is written in Chinese and does not offer any language choice or indicate that the skill is intentionally limited to Chinese-speaking users for a region-specific purpose. Under the policy, forcing a specific language or locale without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The audit report contains materially inconsistent statements about whether P1/P2 remediation was already completed. In a compliance or deployment workflow, this can cause reviewers to rely on an inaccurate security posture and approve or ship a skill under false assumptions about test coverage and documentation state.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The deployment section says all recommendations are closed and the decision is GO, but later text says some items are only suggestions pending confirmation. This creates a decision-integrity problem: downstream operators may treat the artifact as fully remediated when important controls, such as evals or file-hygiene work, may still be outstanding.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The file presents all operational guidance exclusively in Chinese and uses mandatory language such as '必须触发' and '必须包含', with no indication that users may choose another language or locale. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless clearly justified, which is not documented here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The metadata explicitly defines the corpus as Chinese social media posts and the rest of the file provides only Chinese-language templates and style constraints. This creates a language-specific requirement without any indication that users may opt into another language or locale, which matches the policy category for forced language/locale behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The title and content prescribe a Chinese-language writing style throughout, with no indication that the user may choose another language or locale. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless the constraint is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The document is written entirely in Chinese and repeatedly instructs generation and evaluation in that language/style context, but it does not state that the skill is region-specific or provide any opt-in for language selection. Under the policy rule, forcing a specific language without user choice is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The README states the skill learns a distinctive Chinese internet narrative style, which effectively fixes the output locale/language. Because the document does not offer an opt-in choice for other languages or explain a necessary regional limitation, this is a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This markdown file uses Chinese throughout, including headings and instructional examples, and does not indicate that users may choose another language or that the skill is intentionally limited to a Chinese-language context. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.