Back to skill

Security audit

evolver

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because a scanner flagged a shell command risk and raised a concern about processing session-derived material with an external runner; several scripts were only partially inspected.

Consider the shell handling and incomplete inspection before installing. If you use it, avoid processing transcript filenames that an untrusted party can control, and review or redact session material before sending it to an external runner. The scanners do not establish malicious intent, but their coverage is incomplete.

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/evolve-cycle.js
Finding

Automatically selected transcript filenames are interpreted by a shell

Content
View full analysis

Impact

An attacker able to supply the selected transcript filename can execute arbitrary commands as the user running the skill, potentially reading credentials or modifying files outside the authorized transcript-processing operation.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 31)May include surrounding context.

md
1. **Convert** — `sessions-to-transcripts.js` turns OpenClaw session JSONL into `*.transcript.jsonl` (OpenAI chat-completions shape) in `memory/evolution/`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 39)May include surrounding context.

md
- `scripts/sessions-to-transcripts.js` — Session JSONL → generic-chat transcript converter

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 32)May include surrounding context.

md
2. **Ingest** — `evolve-cycle.js` finds the latest transcript and runs `evolver ingest` to create material with extracted signals.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 40)May include surrounding context.

md
- `scripts/evolve-cycle.js` — V2 material-driven cycle runner (`--distill` for no-runner mode)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 35)May include surrounding context.

md
5. **Daemon** — `autoexec-manage.js` starts/stops the autoexec daemon (config at `~/.evomap/autoexec/config.json`, runner e.g. `gemini`).

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

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

md
- `scripts/autoexec-manage.js` — Start/stop/status for the autoexec daemon

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 50)May include surrounding context.

md
- The `--loop` daemon historically doesn't work reliably; use `autoexec-manage.js` instead.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 42)May include surrounding context.

md
- `scripts/verify-gep.js` — Quick GEP prompt verification

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 · SKILL.md (reported line 47)May include surrounding context.

md
## Notes

- V1 commands `evolver run` / `evolver solidify` are **deprecated** in 2.x — use the V2 flow above.
- `evolver solidify` does a `git reset --hard` on failure — commit your work first if you must use it.
- Materials should come from **actual coding work logs**, not QQ chat conversations (chat-derived genes are homogeneous and unproven).
- `evolver cycle` requires a runner (e.g. `gemini`); if the execute queue is disabled, prefer `--distill`.
- The `--loop` daemon historically doesn't work reliably; use `autoexec-manage.js` instead.

Unvalidated Output Injection

High
Category
Output Handling
Confidence
100% confidence
Finding

Model output is used without validation or sanitization. Unvalidated output injected into downstream contexts (SQL, shell, HTML) enables injection attacks and arbitrary code execution.

Content

Scanner excerpt · scripts/detect-cli-changes.js (reported line 36)May include surrounding context.

js
// Extract flags from help text
  const flagRegex = /-\s+(-{1,2}[a-zA-Z][a-zA-Z0-9-]*)(?:\s+\[?[^\]]*\]?)?/g;
  let match;
  while ((match = flagRegex.exec(output)) !== null) {
    const flag = match[1];
    if (output.includes('deprecated') || output.includes('removed')) {
      // Check context for deprecation

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The session-to-cycle workflow lacks a privacy warning about processing session-derived material with an external runner such as Gemini.

Medium
Category
Output Handling
Confidence
85% confidence
Finding

The documented workflow converts actual coding sessions into material for a cycle that can use an external runner such as Gemini, without describing redaction or informed approval before external processing. Coding sessions can contain credentials, proprietary code, and private communications, making this a meaningful disclosure risk, although the supplied documentation does not prove what a particular runner transmits.

Content

No source excerpt is available for this finding.