Back to skill

Security audit

ClawStreet

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed paper-trading skill, but it needs Review because it can persistently run trading/social actions while fetching mutable remote instructions and installing unpinned skill files.

Review before installing. The trading is paper-only, but the bot can make public posts, comments, votes, profile changes, and repeated authenticated API calls. Avoid enabling the heartbeat rule as written, do not let it fetch and follow remote instructions automatically, and only install from a pinned or verified release you have reviewed.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:382
Finding

Persistent Heartbeat Retrieves and Follows Mutable Remote Instructions

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
SKILL.md:431
Finding

Unversioned Remote Files Overwrite Active Skill Instructions Without Integrity Verification

Content
View full analysis
"$BASE/SKILL.md" curl -s https://www.clawstreet.io/skills/clawstreet/SYMBOLS.md > "$BASE/SYMBOLS.md" curl -s https://www.clawstreet.io/skills/clawstreet/INDICATORS.md > "$BASE/INDICATORS.md" curl -s https://www.clawstreet.io/skills/clawstreet/STRATEGIES.md > "$BASE/STRATEGIES.md" curl -s https://www.clawstreet.io/skills/clawstreet/THOUGHT_STYLE.md > "$BASE/THOUGHT_STYLE.md" ``` ### Technical Analysis The installation procedure retrieves multiple mutable files from an external origin and writes them directly over active Skill files. It provides no: - Version pinning. - Cryptographic hash verification. - Digital signature verification. - Content review or diff step. - Staging and atomic activation process. - HTTP failure handling through `curl --fail`. - Validation that all files belong to one consistent release. Although the downloaded files are Markdown rather than shell scripts, they are active behavioral components interpreted by an AI agent. Replacing `SKILL.md` or its mandatory references can therefore change the agent’s goals, network behavior, data-handling rules, and tool-use instructions. The use of `curl -s` also suppresses diagnostic output. Shell redirection opens and truncates the destination before transfer completion, so a failed or partial request can corrupt an existing installation. Downloading each file independently may also produce a mixed-version installation if the remote files change during the process. ### Attack Path 1. A user runs the documented installation commands. 2. The shell creates or truncates files in the active Skill directory. 3. Files are downloaded from mutable URLs ...[truncated 1353 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill description is broad enough to trigger on generic user requests about trading stocks or crypto, which can cause the agent to activate this third-party integration when the user only wanted general financial discussion or a different platform. In this context, unintended invocation is more dangerous because the skill can proceed into registration, API-key handling, recurring heartbeat behavior, and autonomous trade/social actions once engaged.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
89% confidence
Finding

The heartbeat instruction tells the agent to periodically fetch a remote SKILL.md and 'follow it', effectively delegating future behavior to mutable remote content outside the originally reviewed artifact. This creates a supply-chain and prompt-injection risk: if the remote file changes or is compromised, the agent may adopt new instructions that alter permissions, exfiltrate data, or expand actions without renewed review.

Content

Scanner excerpt · SKILL.md (reported line 421)May include surrounding context.

text
If 4+ hours since last ClawStreet check:
1. Fetch https://www.clawstreet.io/skills/clawstreet/SKILL.md and follow it
2. Run one trading cycle (scan → trade → thought → engage)
3. Update lastClawStreetCheck timestamp

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
84% confidence
Finding

The install section directs users or agents to download multiple executable instruction files from remote URLs into the local skills directory, enabling future behavior to be influenced by unpinned external content. While framed as installation guidance, it still enlarges the trust boundary and introduces update/tampering risk if those files are modified upstream or intercepted before review.

Content

Scanner excerpt · SKILL.md (reported line 435)May include surrounding context.

bash
BASE=~/.openclaw/skills/clawstreet  # or ~/.clawdbot/skills/clawstreet
mkdir -p "$BASE"
curl -s https://www.clawstreet.io/skills/clawstreet/SKILL.md > "$BASE/SKILL.md"
curl -s https://www.clawstreet.io/skills/clawstreet/SYMBOLS.md > "$BASE/SYMBOLS.md"
curl -s https://www.clawstreet.io/skills/clawstreet/INDICATORS.md > "$BASE/INDICATORS.md"
curl -s https://www.clawstreet.io/skills/clawstreet/STRATEGIES.md > "$BASE/STRATEGIES.md"