Back to skill

Security audit

LYGO Champion: arkos celestial architect

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed legacy persona helper, but it tells users to run an unpinned remote installer for another skill.

Read this as a legacy persona pack, not as an active automation package. The local files look low-risk, but only run the migration command if you trust the `clawhub` npm package, the DeepSeekOracle publisher, and the current contents of the successor skill; prefer a pinned or reviewed installer/version where possible.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:9
Finding
Mutable Third-Party Installer Creates Supply-Chain Risk## Vulnerability Details **File Location**: `SKILL.md`, line 9 **Vulnerability Type**: `T08: Insecure Dependencies` **Risk Level**: Medium **Vulnerable Code Snippet**: ```markdown > **Consolidated (Δ9 v2):** New installs → `lygo-champion-council`. This slug is legacy retention only. > `npx clawhub@latest install deepseekoracle/lygo-champion-council` ``` ### Technical Analysis The migration instructions recommend executing `clawhub` through `npx` using the mutable `latest` tag. This resolves and runs a third-party npm package version at execution time rather than a version reviewed alongside this project. The command also installs a separately maintained successor Skill whose contents are not present in the audited package. Consequently, the effective installation behavior can change after this project has been reviewed. Compromise of the package publisher, registry entry, distribution channel, or successor Skill could expose users to altered installer code or malicious Skill content. The audited project does not establish that the current upstream package is malicious; the vulnerability is the unpinned trust relationship and unsafe dependency-execution pattern. ### Attack Path 1. An attacker compromises the `clawhub` publishing account, npm package, registry resolution path, or successor Skill distribution account. 2. The attacker publishes a modified release that becomes the target of the mutable `latest` tag or replaces the successor Skill content. 3. A user follows the migration command from `SKILL.md`. 4. `npx` downloads and executes the unreviewed package release. 5. The compromised installer or installed Skill performs attacker-controlled actions with the permissions of the invoking user. ### Impact Assessment Successful exploitation could execute attacker-controlled installer behavior under the invoking user's account. The resulting scope depends on the runtime environment and user permissions, potential ...[truncated 313 chars]
Remediation
## Remediation Suggestions - Replace `clawhub@latest` with an exact, reviewed package version. - Pin the successor Skill to an immutable version, release digest, or cryptographic content hash. - Publish the expected SHA-256 digest and require verification before installation. - Prefer a lockfile-backed installation workflow with integrity metadata. - Document the exact package source and trusted publisher identity. - Avoid directly executing remotely resolved packages where possible; download, verify, inspect, and then run the installer. - Re-audit both the pinned installer version and the successor Skill before recommending them to users. - Add a warning that the migration command executes third-party code with the invoking user's permissions.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The declared description says this is a deprecated slug whose purpose is simply to indicate users should use another skill (lygo-champion-council). The provided code does something materially different: it implements a local validation/self-check utility for a champion pack by reading and validating several repository files. That is an active operational capability not reflected in the description. There are no suspicious external actions or extra permissions, but the primary purpose clearly does not match the declared purpose.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The declared description says this is a deprecated legacy slug whose purpose is to direct users to another skill, with no triggers or permissions and no substantive functional behavior implied. The code, however, performs an actual operation: it locates references/canon.json, reads and parses it, and prints a specific hash field. That is a materially different primary purpose from a deprecation marker or alias, so this should be flagged as a mismatch.

Lp3

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding
Without declared permissions the skill's intent is opaque and cannot be validated.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding
The installation command uses `npx clawhub@latest`, which pulls the newest package version at execution time instead of a pinned, reviewed release. This creates a supply-chain risk: a compromised or breaking upstream release could be executed by users who follow the instruction.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The activation phrases are broad natural-language prompts like “Mint this Champion pack” and “Backfill anchors,” with no explicit confirmation, scoping, or validation requirements. In an agent setting, this can cause unintended invocation on attacker-supplied content or paths, leading the skill to process untrusted inputs and potentially publish or persist misleading ledger or anchor data.

Static analysis

No suspicious patterns detected.