ClawHub
Back to skill

Security audit

Pilot Thread

Security checks for vulnerabilities and agentic risk

Overview

The skill’s purpose is coherent, but its Bash messaging examples use unsafe JSON construction that could expose users to command or payload injection if adapted with untrusted message text.

Review the command examples before installing. Use this only if you trust the Pilot Protocol setup and ensure agents serialize message payloads with a tool such as jq rather than inserting peer, subject, or message text directly into Bash command strings.

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
SKILL.md:48
Finding

Unsafe Shell and JSON Interpolation in Messaging Commands

Content
View full analysis
"$TOPIC" --data "{\"action\":\"create\",\"thread_id\":\"$THREAD_ID\",\"subject\":\"Discussion Topic\"}" ``` ```bash pilotctl --json publish "thread_$THREAD_ID" --data "{\"action\":\"reply\",\"thread_id\":\"$THREAD_ID\",\"message\":\"Response text\"}" ``` The workflow repeats the same construction pattern with variables: ```bash pilotctl --json publish "$PEER" "$TOPIC" --data "{\"action\":\"create\",\"thread_id\":\"$THREAD_ID\",\"subject\":\"$SUBJECT\",\"message\":\"Let's plan the Q2 data processing pipeline.\"}" ``` ```bash pilotctl --json publish "$PEER" "$TOPIC" --data "{\"action\":\"reply\",\"thread_id\":\"$THREAD_ID\",\"message\":\"Great! Can you handle 100K records per hour?\"}" ``` ### Technical Analysis The documented commands construct JSON by embedding values directly into a double-quoted shell argument. They do not use a JSON serializer to escape quotation marks, backslashes, control characters, or other metacharacters. If an agent adapts these templates by inserting attacker-controlled subjects or message text directly into the generated shell command, crafted content can terminate the intended JSON or shell quoting context. Shell syntax introduced before command parsing, including command substitution such as `$(command)` or backticks, can then be evaluated locally. Even when shell execution is not achieved, embedded quotation marks or backslashes can corrupt the JSON payload or alter its structure. The literal examples are benign, but the unsafe pattern is intended for sending conversation content, which may originate from untrusted peers. The vulnerable boundary is therefore the conversion of message content into shell source rather than treating that content strictly ...[truncated 1335 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • 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)

Memory Manipulation

High
Category
Memory Poisoning
Confidence
80% confidence
Finding

Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Content

Scanner excerpt · SKILL.md (reported line 38)May include surrounding context.

md
# pilot-thread

Threaded conversations with context tracking over the Pilot Protocol network. This skill enables organized multi-turn dialogues by maintaining conversation threads, allowing multiple simultaneous discussions with clear context separation and complete message history.

## Commands

Static analysis

No suspicious patterns detected.