Back to skill

Security audit

meta-social-collaboration-mind

Security checks for vulnerabilities and agentic risk

Overview

The skill is not overtly malicious, but it asks agents to keep and reuse local memory and even rewrite its own instructions while overstating unimplemented verification features.

Install only if you are comfortable with a skill that keeps local cross-session notes/preferences and whose documentation encourages future instruction changes. Treat learned_patterns.json as untrusted data, review or clear it regularly, and do not allow automatic edits to SKILL.md without explicit human review.

Vulnerability Patterns
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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)

T02 · Agent Memory Poisoning

Warning
Location
scripts/learner.py:6
Finding

Unvalidated Cross-Session Storage Can Poison Agent Memory

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:49-52, SKILL.md:62-73; persistence implementation in scripts/learner.py:6-17
Vulnerability Type: Persistent storage of untrusted behavioral data
Risk Level: Medium

The skill documentation instructs the agent to retain outcomes, errors, improvement suggestions, and user preferences in learned_patterns.json. It further directs future invocations to apply important stored preferences automatically and recommends writing derived rules back into SKILL.md after repeated failures.

The underlying persistence mechanism accepts arbitrary note text and stores it without validation:

python
MEM = os.path.join(os.path.dirname(__file__), "learned_patterns.json")
def load():
    try:
        return json.load(open(MEM, encoding="utf-8"))
    except Exception:
        return {"total_ops": 0, "total_fails": 0, "patterns": []}
def record(op_ok=True, note=""):
    d = load(); d["total_ops"] += 1
    if not op_ok:
        d["total_fails"] += 1
    if note:
        d.setdefault("patterns", []).append(note)
    json.dump(d, open(MEM, "w", encoding="utf-8"), ensure_ascii=False, indent=2)
    return d

Technical Analysis

record() appends caller-controlled note content directly to persistent JSON storage. There is no schema validation, length restriction, provenance tracking, trust designation, user or project isolation, or approval gate.

Persistent storage alone is not instruction execution. The security risk arises from combining this storage mechanism with the documented workflow in SKILL.md, which tells the agent to reuse stored preferences in later invocations and to convert accumulated experience into modifications of the skill instruction file. If stored natural-language data is subsequently interpreted as trusted behavioral guidance, an attacker can use notes, preferences, or failure descriptions as a durable indirect-instruction channel.

...[truncated 1965 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove automatic modification of SKILL.md; skill instructions should remain immutable during normal execution.
  2. Require explicit, informed user approval before storing preferences, errors, notes, or other cross-session data.
  3. Define a strict JSON schema with permitted keys, value types, maximum lengths, and enumerated preference values.
  4. Treat all persisted natural-language content as untrusted data and never concatenate or promote it into agent instructions.
  5. Separate factual metrics from free-form text. Prefer counters and predefined identifiers over arbitrary notes.
  6. Namespace memory by user and project to prevent one context from influencing another.
  7. Record provenance and timestamps for every stored entry, and provide review, deletion, reset, and expiration controls.
  8. Use atomic file replacement and restrictive filesystem permissions for the memory file.
  9. Align the documentation with the implementation by removing unsupported subcommands or implementing them with validation and approval controls.
  10. If reflection is required, produce proposed changes for human review rather than applying them automatically.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

This is another description-behavior mismatch at the same location: the file presents a meta-skill with advanced oversight and learning claims, but the effective behavior appears to be ordinary collaboration guidance plus logging. Such exaggerated capability claims can mislead downstream agents into delegating sensitive decisions under a false assumption of verification or bounded autonomy.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

This is another description-behavior mismatch at the same location: the file presents a meta-skill with advanced oversight and learning claims, but the effective behavior appears to be ordinary collaboration guidance plus logging. Such exaggerated capability claims can mislead downstream agents into delegating sensitive decisions under a false assumption of verification or bounded autonomy.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill instructs use of a learner module that writes persistent state to local files, but the manifest declares no tool scope or permissions. This creates undeclared file-write capability, which weakens reviewability and user consent because operators cannot tell from the manifest that the skill persists data.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill's stated purpose is reasoning/collaboration improvement, but it also declares persistent usage and preference tracking. Collecting and retaining behavioral data outside the core task increases privacy risk and broadens the data footprint without clear necessity.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill describes automatic persistence of user preferences and usage history without a clear privacy warning or consent flow. Silent retention of cross-session data can surprise users and expose personal habits, language preferences, or task history if the local store is accessed or reused inappropriately.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

Persistent collection of user preferences is not clearly necessary for a social-collaboration meta-skill and may capture personal or sensitive information over time. Unjustified preference storage increases privacy exposure and the blast radius of local data compromise.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly directs retention of user preferences and usage history in a persistent memory file across uses. Cross-session statefulness can enable profiling, accidental storage of sensitive data, and hidden influence on future outputs if users are unaware of or cannot control the memory.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The documentation instructs the system to modify SKILL.md itself as part of iterative learning. Self-modifying operational instructions are risky because they can create integrity drift, bypass review processes, and allow prompt/configuration changes to accumulate without controlled validation.

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 report is written in Chinese, and there is no indication that the skill offers a language/locale choice or that Chinese is required for a region-specific purpose. Under the policy, a skill or related artifact that implicitly forces a specific language without opt-in 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
93% confidence
Finding

The module docstring is written entirely in Chinese and states the skill purpose only in that language, which indicates a language-specific presentation without offering user choice or opt-in. The policy requires avoiding forced language or locale constraints unless explicitly justified or optional.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This file contains natural-language instructions and outputs exclusively in Chinese, including the module docstring and explanatory text. Because the skill does not provide user opt-in or indicate that it is intentionally limited to a Chinese-speaking context, it may violate language/locale policy requirements.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

L48 says '每次使用后自动复盘、积累经验,无需人工维护', implying automatic post-use learning. But L53-L66 instruct users to manually call Bash commands such as 'record', 'prefer', 'insight', and 'reflect', which contradicts the claim of automatic operation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The example command stores 输出语言 --val 中文, which presents Chinese as the default output language preference. Because the file does not explicitly offer language choice or frame this as an optional user-selected example, it can be read as imposing a specific language/locale without opt-in.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This code writes persistent data to learned_patterns.json using json.dump, but there is no confirmation prompt, user-facing log message, or explanatory comment at the write site. For a code file, persistent file modification is a safety-relevant operation that should include some visible disclosure unless clearly communicated elsewhere.

Content

No source excerpt is available for this finding.