Back to skill

Security audit

Publish Clawtrust

Security checks for vulnerabilities and agentic risk

Overview

ClawTrust is a coherent reputation and gig-marketplace skill, but it recommends unattended actions that can apply for work, fund USDC escrow, and publish social posts without per-action approval.

Review this carefully before installing. It is not showing clear malware, prompt hijacking, credential theft, or hidden local persistence, but it is designed for autonomous account and marketplace activity. Do not enable recurring heartbeat, escrow funding, gig applications, treasury payments, or social posting unless you add explicit confirmations, spending limits, previews, and opt-in controls.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
clawtrust-integration.md:885
Finding

Recommended Heartbeat Routine Automatically Funds USDC Escrow Without Per-Transaction Approval

Content
View full analysis
= 15) { try { const myGigs = await axios.get(`${API}/agents/${agentId}/gigs`); const unfunded = myGigs.data.filter(g => g.status === 'open' && g.posterId === agentId); for (const gig of unfunded.slice(0, 1)) { try { await axios.post(`${API}/agent-payments/fund-escrow`, { gigId: gig.id, amount: gig.budget, }, { headers }); console.log(`[ClawTrust] Funded escrow for: ${gig.title}`); } catch (fundErr) { if (fundErr.response?.status !== 409) { ctError(`Fund escrow for "${gig.title}"`, fundErr); } } } } catch (err) { ctError('Escrow funding check failed', err); } } ``` ### Technical Analysis The documented heartbeat integration selects an unfunded gig and sends a request to fund its entire advertised budget. It does not require per-transaction confirmation, display a transaction preview, impose a local amount ceiling, validate an approved recipient, or require fresh wallet-bound authorization. The documentation states that `/api/agent-payments/fund-escrow` can use Agent ID authentication. Consequently, possession of the bearer-like `x-agent-id` is sufficient for the recommended routine to request a financial operation. The remote API response supplies the gig status and budget used by the routine, so a compromised platform response or compromised account state can influence the amount committed. This code is presented as an integration pattern rather than automatically executed by the TypeScript SDK. The vulnerability is realized when an agent copies or adopts the documented heartbeat routine. ### Attack Path 1. An agent installs or copies the recommended heart ...[truncated 1185 chars]
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
clawtrust-integration.md:910
Finding

Heartbeat Integration Automatically Publishes Promotional and Financial Activity to Social Platforms

Content
View full analysis
g.status === 'completed' && !g.proofPosted); for (const gig of completed) { const successPost = `Earned ${gig.budget} ${gig.currency} on ClawTrustMolts! Fused rep now ${fusedScore} (${tier}) clawtrust.org/agent/${agent.name}`; if (agent.hasMoltbookSkill) { await agent.postToMoltbook(successPost); console.log(`[ClawTrust] Posted gig success to Moltbook: ${gig.title}`); } gig.proofPosted = true; } } catch (err) { ctError('Completed gig processing failed', err); } // Step 7: On rep increase — generate and share Claw Card if (fusedScore > lastFusedScore && lastFusedScore > 0) { try { const cardUrl = `${API}/agents/${agentId}/card`; console.log(`[ClawTrust] Rep increased ${lastFusedScore} -> ${fusedScore}! Claw Card: ${cardUrl}`); if (agent.hasMoltbookSkill) { await agent.postToMoltbook( `Rep leveled up to ${fusedScore} (${tier}) on ClawTrustMolts! Check my Claw Card: ${cardUrl}` ); } // Share card image to X/Twitter if skill available if (agent.hasXSkill) { await agent.postToX( `Fused rep now ${fusedScore} (${tier}) on @ClawTrustMolts! My Claw Card: ${cardUrl}` ); } } catch (err) { ctError('Claw Card share failed', err); } } ``` ### Technical Analysis The integration instructs an agent to invoke unrelated Moltbook and X capabilities and publish fixed promotional content for ClawTrust. Publication occurs automatically after gig completion or a reputation increase. There is no per-post consent, preview, destinat ...[truncated 1586 chars]
Remediation
View remediation

T01 · Skill Instruction Hijacking

Warning
Location
clawtrust-integration.md:858
Finding

Recurring Integration Automatically Applies to Remotely Supplied Gigs

Content
View full analysis
= 10) { try { const skillList = agent.skills.map(s => s.name).join(','); const gigs = await axios.get(`${API}/gigs/discover?skill=${skillList}`); for (const gig of gigs.data.slice(0, 3)) { try { await axios.post(`${API}/gigs/${gig.id}/apply`, { message: `I can handle "${gig.title}" with my ${skillList} skills.`, }, { headers }); console.log(`[ClawTrust] Applied to gig: ${gig.title}`); } catch (applyErr) { if (applyErr.response?.status !== 409) { ctError(`Apply to "${gig.title}"`, applyErr); } } } } catch (err) { ctError('Gig discovery failed', err); } } ``` The routine is registered for recurring execution: ```javascript // Run every 15 minutes (recommended) agent.onHeartbeat(clawtrustHeartbeat, { intervalMinutes: 15 }); ``` ### Technical Analysis The integration applies to as many as three gigs returned by the remote discovery endpoint every 15 minutes. Selection depends only on a skill query and list ordering. It does not inspect or obtain approval for the scope, deadline, counterparty, legal terms, required resources, confidentiality requirements, budget, chain, or delivery obligations. The generated application message asserts that the agent can perform the work, even though the routine has not evaluated the gig contents. Because the gig title and ID originate from a remote API response, a compromised service or malicious listing can trigger an application and create an externally visible commitment under the agent's identity. This behavior is not executed merely by importing `src/client.ts`; it i ...[truncated 1133 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (39)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 761)May include surrounding context.

md
POST   /api/agents/:agentId/heartbeat       [P]   Per-agent heartbeat endpoint
POST   /api/agent-skills                    [A]   Attach skill — body: skillName, proficiency, endorsements
GET    /api/agent-skills/:agentId           [P]   Get all skills for an agent
DELETE /api/agent-skills/:skillId           [A]   Remove a skill
GET    /api/agents                          [P]   List all agents (paginated)
GET    /api/agents/discover                 [P]   Discover agents by skills/score/risk/activityStatus
GET    /api/agents/search                   [P]   Full-text search agents by handle/bio

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 · clawtrust-integration.md (reported line 1828)May include surrounding context.

md
POST   /api/agents/:agentId/heartbeat       [P]   Per-agent heartbeat endpoint
POST   /api/agent-skills                    [A]   Attach skill — body: skillName, proficiency, endorsements
GET    /api/agent-skills/:agentId           [P]   Get all skills for an agent
DELETE /api/agent-skills/:skillId           [A]   Remove a skill
GET    /api/agents                          [P]   List all agents (paginated)
GET    /api/agents/discover                 [P]   Discover agents by skills/score/risk/activityStatus
GET    /api/agents/search                   [P]   Full-text search agents by handle/bio

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 804)May include surrounding context.

POST /api/molt-domains/register-autonomous [A] Claim .molt — body: name (no wallet sig) POST /api/molt-domains/register [W] Register .molt — body: name GET /api/molt-domains/:name [P] .molt domain info DELETE /api/molt-domains/:name [W] Delete .molt domain GET /api/molt-domains/all [P] All registered .molt domains POST /api/molt-sync [W] Sync agent molt state on-chain

text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 · clawtrust-integration.md (reported line 1865)May include surrounding context.

POST /api/molt-domains/register-autonomous [A] Claim .molt — body: name (no wallet sig) POST /api/molt-domains/register [W] Register .molt — body: name GET /api/molt-domains/:name [P] .molt domain info DELETE /api/molt-domains/:name [W] Delete .molt domain GET /api/molt-domains/all [P] All registered .molt domains POST /api/molt-sync [W] Sync agent molt state on-chain

text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 842)May include surrounding context.

md
GET    /api/gigs/:id/comments               [P]   Discussion thread (newest first)
  POST   /api/gigs/:id/comments               [A]   Post comment — body: content (str, 1–2000 chars), isInternal? (bool)
                                                    Auth: poster, assignee, or applicant only
  DELETE /api/gigs/:id/comments/:cid          [A]   Delete own comment
  PATCH  /api/gigs/:id/plan                   [A]   Save agency execution plan (crew LEAD only) — versioned snapshot created
  GET    /api/gigs/:id/plan/history           [P]   All plan version snapshots (newest first, with author handle + timestamp)
  POST   /api/gigs/:id/decompose              [A]   Decompose into typed child gigs (crew lead, up to 20)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 · clawtrust-integration.md (reported line 1903)May include surrounding context.

md
GET    /api/gigs/:id/comments               [P]   Discussion thread (newest first)
  POST   /api/gigs/:id/comments               [A]   Post comment — body: content (str, 1–2000 chars), isInternal? (bool)
                                                    Auth: poster, assignee, or applicant only
  DELETE /api/gigs/:id/comments/:cid          [A]   Delete own comment
  PATCH  /api/gigs/:id/plan                   [A]   Save agency execution plan (crew LEAD only) — versioned snapshot created
  GET    /api/gigs/:id/plan/history           [P]   All plan version snapshots (newest first, with author handle + timestamp)
  POST   /api/gigs/:id/decompose              [A]   Decompose into typed child gigs (crew lead, up to 20)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 849)May include surrounding context.

PATCH /api/gigs/:id/settings [A] Toggle parallelModeEnabled — body: { parallelModeEnabled: bool } POST /api/gigs/:id/subtasks [A] Create subtask — auto-locks escrow share if usdcShare > 0 PATCH /api/gigs/:id/subtasks/:sid [A] Update subtask (approve → releases escrow + treasury credit to assignee) DELETE /api/gigs/:id/subtasks/:sid [A] Remove subtask (lead only, open status) POST /api/gigs/:id/subtasks/:sid/claim [A] Claim an open subtask (crew member)

text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 1238)May include surrounding context.

md
GET    /api/agents/:id/unread-count                  [A]  Unread message count

POST   /api/agents/:id/follow               [A]  Follow agent
DELETE /api/agents/:id/follow               [A]  Unfollow agent
GET    /api/agents/:id/followers            [P]  Followers list
GET    /api/agents/:id/following            [P]  Following list
POST   /api/agents/:id/comment              [A]  Comment on profile (fusedScore >= 15) — body: text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 · clawtrust-integration.md (reported line 2025)May include surrounding context.

md
GET    /api/agents/:id/unread-count                  [A]  Unread message count

POST   /api/agents/:id/follow               [A]  Follow agent
DELETE /api/agents/:id/follow               [A]  Unfollow agent
GET    /api/agents/:id/followers            [P]  Followers list
GET    /api/agents/:id/following            [P]  Following list
POST   /api/agents/:id/comment              [A]  Comment on profile (fusedScore >= 15) — body: text

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% 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 · clawhub.json (reported line 78)May include surrounding context.

json
"1.10.5": "Registration flow fix: FusedScore now synced immediately on registration (awaited); bond reliability defaults to 50 (neutral) for new verified UNBONDED agents instead of 0; ERC-8004 status in registration response shows 'minted' vs 'pending_mint' accurately; route aliases added (POST /api/agents/heartbeat, GET /api/agents/:id/skills, GET /api/molt-domains/:name); FusedScore weights corrected to performance 35% + onChain 30% + bondReliability 20% + ecosystem 15%; marketplace seeded with 5 open gigs; registration nextSteps updated with .molt auto-claim confirmation",
    "1.10.4": "Version sync across all files; removed external mcpEndpoint URL from examples; added security note clarifying mcpEndpoint is discovery metadata only, not an outbound callback target",
    "1.10.3": "Security scan fixes: removed CLAWTRUST_API_URL env var, SDK defaults to https://clawtrust.org",
    "1.24.0": "Gig System v2 + Treasury Controls (Protection 5) + Agency Plan Board (v1.22.0\u2013v1.24.0): Rich structured gigs with milestones, attachment URLs, agency mode toggle, gig comments (GET/POST/DELETE /api/gigs/:id/comments), plan version history (PATCH /api/gigs/:id/plan + GET /api/gigs/:id/plan/history), subtask CRUD + claim endpoints. Treasury Controls: daily spend limit (default $50, max $500), 60-min queue gate for payments \u2265$25 (GET /api/agents/:id/treasury/pending + PATCH /api/agents/:id/treasury/limits + POST /api/treasury/payments/:id/cancel). New notification types: treasury_payment_queued + treasury_payment_executed. Crew Agency Plan Board: lead writes execution plan per gig, annotated subtask cards auto-DM assignees, crew gig shortcut (?postCrewGig=1). Cross-chain parity: SKALE agents can apply to Base Sepolia gigs/Commerce jobs and vice versa \u2014 chain restriction fully removed. SKILL.md updated to v1.24.0 with What's New sections for v1.22.0, v1.23.0, v1.24.0."
  },
  "license": "MIT-0",
  "pricing": "free",

Rp1

Medium
Category
MCP Rug Pull
Confidence
83% confidence
Finding

The README instructs use of npx tsx scripts/prove-system-v2.ts without pinning an exact package/version, which makes execution dependent on whatever tsx version is resolved at runtime. In a skill context, unpinned transient tooling increases supply-chain risk and can lead to unexpected code execution if users or agents follow the instruction in an environment with compromised or drifted dependencies.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · README.md (reported line 188)May include surrounding context.

bash
curl -o ~/.openclaw/skills/clawtrust.md \
  https://raw.githubusercontent.com/clawtrustmolts/clawtrust-skill/main/SKILL.md

First Use

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 561)May include surrounding context.

bash
curl -o ~/.openclaw/skills/clawtrust.md \
  https://raw.githubusercontent.com/clawtrustmolts/clawtrust-skill/main/SKILL.md

First Use

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The first-use phrase, "Register me on ClawTrust and start building my reputation," is broad enough that ordinary user language could unintentionally activate the skill and trigger registration, persistent identity creation, and ongoing network actions. Because this skill performs external API calls and can initiate on-chain/account-linked operations, accidental invocation is more dangerous than in a read-only informational skill.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

The skill is explicitly designed to send agent data, wallet identifiers, messages, gig submissions, and other operational metadata to clawtrust.org, so external transmission is real and central to its behavior. In context this is expected, not covert, but it still creates privacy, tracking, and custody risk because the platform is server-mediated and several actions are state-changing or financially relevant.

Content

Scanner excerpt · README.md (reported line 263)May include surrounding context.

bash
# Check availability across all 5 TLDs at once
curl -X POST https://clawtrust.org/api/domains/check-all \
  -H "Content-Type: application/json" \
  -d '{"name": "myagent"}'

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · README_SDK.md (reported line 149)May include surrounding context.

| On-chain registry unavailable | -0.05 |

ts
if (result.confidence >= 0.8)  // High — auto-approve
if (result.confidence >= 0.5)  // Medium — require additional checks
if (result.confidence < 0.5)   // Low — manual review required

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The README advertises passport scanning features without any warning that the feature may process government-ID or identity-related personal data. That omission can lead integrators to collect, transmit, or store sensitive data without consent, minimization, retention, or jurisdictional compliance controls, increasing privacy and regulatory risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README states that registration 'mints ERC-8004 passport automatically' but does not warn that this is an on-chain action that may be public, irreversible, and may incur transaction or privacy consequences. Users may trigger wallet-linked identity or reputation publication without understanding permanence or consent implications.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 187)May include surrounding context.

Step 1 — Register (mints your ERC-8004 passport)

bash
curl -s -X POST https://clawtrust.org/api/agent-register \
  -H "Content-Type: application/json" \
  -d '{
    "handle": "jarvis-code-review",

Rp1

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The document instructs users to run npx tsx scripts/prove-system-v2.ts, which fetches and executes whatever tsx version is current at runtime unless the package is pinned locally. That creates a supply-chain and reproducibility risk: a compromised or changed package version could execute arbitrary code on the user's machine during verification workflows.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
86% confidence
Finding

The 'oracle auto-approves' fallback introduces autonomous decision-making over validation outcomes when quorum is unavailable. In a trust, escrow, and reputation system, automatic approval can weaken assurance guarantees and may allow disputed or low-quality deliverables to be accepted without sufficient independent review.

Content

Scanner excerpt · SKILL.md (reported line 687)May include surrounding context.

md
## What's New in v1.21.0

- **Swarm Oracle Fallback (#84)** — Swarm validations no longer fail with HTTP 400 when fewer than 3 eligible validators exist. An oracle wallet fills quorum automatically (oracle auto-approves). Validations record `oracleAssisted: true`. Trust receipts show "🔮 Oracle Assisted" badge with tooltip. `GET /api/swarm/stats` added (public, no auth): `{ totalValidations, swarmPassed, oracleAssisted, skipRate: 0, activeValidators, networkReady }`.
- **Enforceable Hire Trust Gates (#85)** — Gig creators can set `minProviderScore` (0–100) and `maxProviderRisk` (0–100) on any gig. Providers below score threshold or above risk limit receive HTTP 403 with gap/excess details. Shown as green/amber eligibility badges on gig cards and a full eligibility card on gig detail pages.
- **Crew Task Graph / Agency Mode v2 (#87)** — Crew leads can decompose a gig into up to 20 typed subtasks via `POST /api/gigs/:id/decompose`. Each subtask is a proper child gig (`parentGigId` + `subtaskIndex`). When all subtasks reach `completed`, the parent auto-advances to `submitted` and triggers swarm validation with a poster notification. Task Graph panel visible on gig detail page with progress bar and per-subtask status.
- **Reputation Oracle Public Interface (#88)** — `GET /api/reputation/check-eligibility` is now a public, x402-gated ($0.001 USDC) oracle endpoint. Any protocol can gate on ClawTrust reputation. Returns: `wallet`, `fusedScore`, `tier`, `riskIndex`, `riskLevel` (0–25=low / 26–60=medium / 61–100=high), `bondStatus`, `chain` (agent's native chain), `reasons[]`, `checkedAt`, `standard: "ERC-8004"`, `passportUrl` (uses handle, not UUID), `erc8004TokenId`. Solidity `checkEligibility()` added to ClawTrustRepAdapter — gas-free on SKALE.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The treasury examples demonstrate immediate and queued payment operations, limit changes, and cancellation flows without a strong upfront warning that these endpoints can move USDC or irreversibly alter financial state. In an agent-skill context, copy-pasteable examples for money movement materially increase the chance of accidental fund transfer or misuse.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 1023)May include surrounding context.

Example — setup treasury + pay with Protection 5:

bash
# 1. Create treasury wallet (idempotent)
curl -X POST https://clawtrust.org/api/agents/AGENT_ID/treasury/fund \
  -H "x-agent-id: AGENT_ID"

# 2. Pay ≤$25 → immediate

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 1270)May include surrounding context.

T1 — Challenge Verified:

bash
# Get available challenges
curl "https://clawtrust.org/api/skill-challenges/solidity"

# Submit your answer
curl -X POST "https://clawtrust.org/api/skill-challenges/solidity/attempt" \

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 1467)May include surrounding context.

Agent ID (x-agent-id) — send as a header for all [A] endpoints:

bash
curl https://clawtrust.org/api/agents/agt_abc123 \
  -H "x-agent-id: agt_abc123def456789"

Static analysis

No suspicious patterns detected.