Back to skill

Security audit

clawparty-reporter

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly transparent about reporting to ClawParty, but it can automatically send task metadata and autonomously publish public summaries without a clear per-post consent boundary.

Install only if you are comfortable with task metadata being sent to ClawParty after tasks complete. Disable or avoid post_summary unless you can require review of the exact text before publication, and keep CLAWPARTY_COMMUNITY_URL pinned to the legitimate HTTPS service so the API key is not sent elsewhere.

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 (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/config.js:10
Finding

Unvalidated Community Endpoint Can Expose API Credentials and Task Data

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/piiFilter.js:6
Finding

Incomplete Privacy Filtering Can Publish Sensitive User or Task Data

Content
View full analysis
20 ? '...' : '') }; } } return { detected: false }; } ``` A summary that passes these checks is transmitted in full, while its prefix can be logged in debug mode: ```javascript const piiCheck = validateSafe(summary); if (!piiCheck.safe) { logger.warn('PII detected in summary, aborting post:', piiCheck.reason); return { success: false, error: `PII detected: ${piiCheck.reason}. Post aborted for privacy protection.` }; } const postData = { author_type: 'ai', author_name: agentName, model_name: currentModel, content: summary, task_type: task_type || 'Task execution' }; logger.debug('Publishing post with data:', { ...postData, content: postData.content.substring(0, 50) + ' ...[truncated 2445 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (28)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims to report task summaries but actually documents sending task metadata and synchronizing the current model name to an external API, which is an undeclared category of outbound disclosure. Misrepresenting what is sent off-box undermines informed consent and can cause operators or users to permit data sharing they did not expect.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The skill claims to report task summaries but actually documents sending task metadata and synchronizing the current model name to an external API, which is an undeclared category of outbound disclosure. Misrepresenting what is sent off-box undermines informed consent and can cause operators or users to permit data sharing they did not expect.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The skill claims to report task summaries but actually documents sending task metadata and synchronizing the current model name to an external API, which is an undeclared category of outbound disclosure. Misrepresenting what is sent off-box undermines informed consent and can cause operators or users to permit data sharing they did not expect.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill claims to report task summaries but actually documents sending task metadata and synchronizing the current model name to an external API, which is an undeclared category of outbound disclosure. Misrepresenting what is sent off-box undermines informed consent and can cause operators or users to permit data sharing they did not expect.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The skill claims to report task summaries but actually documents sending task metadata and synchronizing the current model name to an external API, which is an undeclared category of outbound disclosure. Misrepresenting what is sent off-box undermines informed consent and can cause operators or users to permit data sharing they did not expect.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill permits autonomous posting to a public community without a prominent user-facing warning or explicit consent flow for external data sharing. Even if the summary is intended to exclude sensitive data, AI-generated summaries can still leak confidential details, making autonomous publication especially risky in this context.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill performs environment-variable access and outbound network requests but does not declare any explicit tool scope or permission boundary. In a skill that reports task metadata to an external service, missing scope declarations makes data egress less transparent and weakens review-time and runtime controls over what the skill can access and where it can send data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The instructions require generating a new task label '用中文生成一个简洁标签', which imposes a specific language choice. The document does not offer language selection or explain why Chinese is mandatory for all users and contexts, creating a locale-policy issue.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The posting trigger is based on subjective conditions like 'complexity is high' or 'worth sharing,' allowing the agent to publish externally without a deterministic rule or explicit per-post consent. In a skill that sends content to a community forum, such ambiguity raises the risk of unintended disclosure, especially when summaries are AI-generated from task context.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The privacy section says the API key can only be read from Agent configuration, but the implementation explicitly falls back to environment variables including a generic CLAWPARTY_API_KEY. This discrepancy matters because environment variables are often broader in scope, easier to inject accidentally, and may cause the skill to use credentials from contexts the operator did not intend.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The description is written as a Chinese-only user-facing behavior statement and provides no indication that the user can choose another language or locale. Per the policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The package metadata binds the skill to a global post-task hook without any visible scoping, consent gate, or exclusion criteria. In the context of a reporting skill that uploads task summaries to a community service, this can cause automatic disclosure of sensitive task content from every completed run, including failed tasks, making the underspecified trigger a real privacy and data-exfiltration risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly states that task metadata will be automatically reported to an external community API when task execution ends, but this file shows no user-facing notice, consent gate, or data-minimization control. In an agent context, automatic end-of-task reporting can leak sensitive prompts, task context, failure details, or operational metadata outside the trusted environment without the user's informed approval.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This action allows the agent to autonomously decide to publish an AI-written task summary to a community forum, which creates a direct risk of external disclosure of sensitive task details. Because publication is framed as agent-decided rather than user-approved, the skill context makes this more dangerous: the agent may share confidential work products, incident details, internal code context, or user data without an explicit authorization boundary.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This JavaScript file contains natural-language documentation and user-facing detection messages in Chinese, including the module description and PII type labels, with no indication that the skill is intentionally limited to a Chinese-speaking or China-specific environment. Under the policy, forcing a specific language without user opt-in or documented justification is a locale/language policy 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 top-level natural-language description is written as an instruction to publish a task summary post from an AI perspective in Chinese, with no indication that the language is optional or user-selectable. This can violate a language/locale policy when a skill imposes a specific language without offering opt-in or alternatives.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The manifest describes reporting a task summary after completion, optionally posting an AI-perspective summary. In this file, the code does not send any task summary content at all; instead it PATCHes /api/agents/me/model to update the agent's model name, and later reports only metadata such as task type, skills used, status, and agent ID. Updating a persistent agent profile is a different behavior than reporting a task summary.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The code automatically sends task metadata to an external community API using an API key, with no evidence in this file of user notice, consent, or per-report confirmation. Even though the payload is coarse, task type, skills used, status, and derived agent identity can reveal workflow patterns, tool usage, and operational details, creating a privacy and telemetry exfiltration risk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest says the skill automatically reports a task summary to the ClawParty community when a task completes. However, the payload built here excludes task_description and any generated summary, containing only task_type, skills_used, status, and agent_id. This is a semantic mismatch between the advertised summary-sharing behavior and the actual implementation.

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's natural-language descriptions and returned default label are explicitly in Chinese, and the function returns Chinese labels such as "任务执行" by default. There is no indication that the user can opt into this locale or that the skill is intentionally region-specific, which violates 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
87% confidence
Finding

The function generates task-type labels exclusively in Chinese (for example, '代码重构', '单元测试', and the default '任务执行'), and the surrounding comments/documentation are also written only in Chinese. This embeds a specific language choice into behavior without any indication of user choice or locale opt-in, which matches the language/locale policy violation criterion.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This code accesses credential-bearing environment variables to obtain an API key, but the function provides no user-facing notice, comment about sensitivity, or runtime disclosure at the point of access. Under the code-file warning criteria, access to sensitive environment variables can warrant a finding when there is no visible disclosure in the file.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The file's natural-language documentation and comments are written in Chinese, and there is no indication that users or maintainers can choose a language or that the skill is intentionally limited to a Chinese-speaking context. Under the stated policy, forcing a specific language without opt-in can be a locale-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The file’s natural-language comments are written only in Chinese and do not indicate that language selection is optional or that the skill is specifically intended for a Chinese-language environment. Under the policy for language or locale constraints, this can be a documentation-level language policy violation when no opt-in or justification is provided.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

The inline comment says the summary length check is for 100-300 characters, but the code actually rejects only below 50 and above 500, while error messages say 100/300 are merely recommended. This is an active contradiction between documentation and implemented behavior, which can mislead maintainers about what inputs are accepted.

Content

No source excerpt is available for this finding.