Back to skill

Security audit

call-pele-answer

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report external transmission of answer content, activity and capability concerns beyond its stated presentation purpose, and incomplete analysis.

Treat this as a review-needed skill and avoid installing it if you handle confidential answer content or need assurance that processing stays local. Before considering installation, clarify whether external classification can be disabled, whether the self-hosted route can be bypassed, and whether activity logging can be limited. Skillspector analyzed only 13 of 45 components fully, so evidence about the remaining behavior 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/core/engine.mjs
Finding

Component admission bypasses the selected self-hosted routing endpoint

Content
View full analysis

Impact

Confidential block contents can be disclosed to an unintended external processor even though the selected self-hosted router succeeds. This finding does not establish theft of unrelated credentials.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (11)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The supplied chunk implements rule registration, not Markdown presentation, visualization selection, or offline HTML generation. Nothing in the chunk connects this rule to the declared purpose. The empty callback does not establish any actual financial analysis, resource access, or other substantive checking behavior.

Content

No source excerpt is available for this finding.

Code capability 'env' detected in scripts/core/cache.mjs but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

Code capability 'network' detected in scripts/core/components.mjs but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'network' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · scripts/vendor/pele.mjs (reported line 40)May include surrounding context.

js
<s;c++){let l=e.charCodeAt(c);if(l===58)a=!0;else if(l===35){if(a){i=c;break}}else Ye(l)&&(a=!1)}if(i!==-1){let c=e.lastIndexOf(";",s-1);c>i&&(r+=e.slice(o,c),o=c+1)}n=s>=e.length?-1:e.indexOf(t,s+1)}return o===0?e:r+e.slice(o)}function nh(e){return e.includes("#")?th(th(e,"style"),"classDef").replace(WS,t=>{let n=t.substring(1,t.length-1);return(US.test(n)?"fl°°":"fl°")+n+"¶ß"}):e}var rh="7q7q9vfgfgopij5b9999atg87q997q7qfgfgfgfgfgfgfgfgfgfg7q7qg8g8g8fgs7ijijk2k2ijgzlmk27qdwijfgn5k2lmijlmk2ijgzk2ijq8ijijgz7q7q7qd1fg99fgfgdwfgfg7qfgfg6666dw66n5fgfgfgfg99dw7qfgdwk2dwdwdw9a789ag8",sh="7q99d6fgfgopk26m9999atg87q997q7qfgfgfgfgfgfgfgfgfgfg9999g8g8g8gzr3k2k2k2k2ijgzlmk27qfgk2gzn5k2lmijlmk2ijgzk2ijq8ijijgz997q99g8fg99fggzfggzfg99gzgz7q7qfg7qopgzgzgzgzatfg99gzfglmfgfgdwat7satg8";var ah=1.2;function ih(e){let t=new Uint16Array(95);for(let n=0;n<95;n++)t[n]=parseInt(e.substr(n*2,2),36);return t}var GS,YS;function HS(e){return e>=4352&&e<=4447||e>=11904&&e<=42191||e>=44032&&e<=55203||e>=637

L37 requires view labels in Chinese without user opt-in.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The assembly step explicitly instructs the agent to change view labels into reader-friendly Chinese. This imposes a specific output language without opt-in or a documented justification, and conflicts with L23's instruction to use the user's language.

Content

No source excerpt is available for this finding.

Profiles agent activity beyond presentation work and recommends reducing unrelated verification.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The advice generator recommends running only pe check before delivery when unrelated shell calls exceed a timing threshold. A presentation checker cannot replace task-specific tests or security verification, so following this recommendation can weaken validation of work outside the skill's scope; the local timing analysis itself is not the primary problem.

Content

No source excerpt is available for this finding.

Persists agent-wide activity metadata unrelated to rendering explanations.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

When installed as an agent-wide hook, this code persistently records session identifiers, tool names, timestamps, and prompt lengths even for activity unrelated to presentation. Content minimization and private-file helpers reduce exposure, but the shown implementation has no collection opt-in, scope filter, or retention limit, creating a limited local privacy risk rather than demonstrated exfiltration.

Content

No source excerpt is available for this finding.

Default routing selects third-party content classification without manifest disclosure.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The default router preset selects external classification and can fall back to a third-party forwarding service, allowing answer blocks to leave the machine during ordinary rendering. SKILL.md does disclose external classification, so the claim of completely absent disclosure is overstated, but broad automatic invocation and opt-out rather than opt-in routing still create a confidentiality risk for sensitive explanations.

Content

No source excerpt is available for this finding.

Uploads answer content to external classification services beyond the manifest's disclosed presentation behavior.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

Both adapter implementations transmit complete supplied text states to external services, with no consent or sensitive-content gate visible at this boundary. Classification supports the presentation workflow and is mentioned in SKILL.md, but producing offline HTML does not make its preparation offline or prevent disclosure of confidential answer content.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

This endpoint is connected to executable POST logic that sends answer text to TypeSafe, so it is not merely an inert URL reference. HTTPS and provider-specific authentication are appropriate, and there is no evidence of credential theft, but confidential answer content can still be disclosed if external processing has not been authorized.

Content

Scanner excerpt · scripts/routers/systemone.mjs (reported line 99)May include surrounding context.

js
// Built-in adapters. Users add their own in config: { "adapters": [{ "name": "my-proxy", "kind": "native",
// "url": "https://…/v1/systemone", "keyEnv": ["PE_KEY_MY_PROXY"] }] } — user config only
export const BUILTIN_ADAPTERS = [
  nativeAdapter({ name: 'typesafe', url: 'https://api.typesafe.ai/v1/systemone', keyEnv: ['TYPESAFE_API_KEY'], describe: 'TypeSafe 直连 System One(需 TYPESAFE_API_KEY)' }),
  nativeAdapter({ name: 'classifier-systemone', url: 'https://classifier.dev/v1/systemone', keyEnv: ['CLASSIFIER_API_KEY', 'CLASSIFY_API_KEY'], anonymousKey: 'unused', keyedConcurrency: 4, describe: 'classifier.dev 转发的 System One(无 key 用 IP 免费额度,4 并发)' }),
  shimAdapter({ name: 'classifier', url: 'https://classifier.dev/v1/classify', keyEnv: ['CLASSIFIER_API_KEY', 'CLASSIFY_API_KEY'], describe: 'classifier.dev 批量分类,按问题批量模拟 System One(无 key 可用)' }),
];

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · scripts/vendor/PELE-LICENSE.txt (reported line 16)May include surrounding context.

text
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,