Back to skill

Security audit

SignalDig Growth Research

Security checks for vulnerabilities and agentic risk

Overview

ClawScan’s verdict is benign with medium confidence: the checks found no issue in the completed AIG scan, while Skillspector reported several findings judged expected for this skill’s purpose; some reference checks were incomplete.

The verdict is benign, but confidence is medium. Skillspector’s scan was partial: although all five listed components were inspected, it could not resolve several local references, so the evidence is incomplete. Consider whether the skill’s instructions fit your needs and review its reference coverage before installing.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 144)May include surrounding context.

md
3` Fashion; `11004` Music & Entertainment; `11005` Games; `11006` Education & Work; `11007` Sports & Outdoor; `11008` Creativity & Talent; `11009` Family & Relationships; `11010` Food & Drink; `11011` Vlog & Selfie; `11012` Society; `11013` Lifestyle & Leisure; `11014` Automotive & Transport; `11015` Technology & Finance. Do not send category names as IDs.
  - `target.tiktok_order_by_metric` defaults to `1` (views); `2` sorts by engagement rate and `3` by six-second view rate. `target.tiktok_organic_only` defaults to `false`. `target.tiktok_page` starts at `1`; `target.tiktok_limit` defaults to `20` and cannot exceed `100`. Read `data.pagination.hasMore` before requesting another page.

Do not add TikTok to topic-search `scope.platforms` or provide account Cookie values or search tracking IDs. Results contain a bounded public projection; video CDN links and access tokens are excluded. Empty App V3 search pages can be valid; inspect the returned `has_more` before continuing.

### Zhihu

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
80% 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 · SKILL.md (reported line 31)May include surrounding context.

md
- Use `research_seo_signals` for keyword demand, intent, related queries, SERP, trends, competitors, GEO visibility, backlinks, ranked keywords, or traffic evidence.
- Use a named Social operation for bounded entity or feed retrieval. Threads topic search belongs to the unified Social Tool.
- When one question genuinely requires both directions, submit one bounded analysis to each tool and keep the two `analysis_id` values separate.
- Ask one short clarification only when choosing Social versus SEO, the required platform set, or the SEO business scope would otherwise be unsafe. Do not ask the user about providers, workflows, bindings, page tokens, operation names, or storage.

## Unified Research Execution Contract

Scope Creep

Low
Category
Excessive Agency
Confidence
75% 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 · SKILL.md (reported line 266)May include surrounding context.

md
- Invalid business scope: explain the missing or invalid user-level field and ask only for that field.
- Deadline reached: return the current `analysis_id` and status so a later turn can resume with `get`; do not submit again.
- `partial`: use available evidence and identify the missing coverage.
- A requested evidence family with no records is a reported coverage limitation, not permission to resubmit, broaden scope, or discard successful families.
- `failed`: say SignalDig could not complete the requested research and suggest retrying later. Never reconstruct technical causes from hidden fields.

Read [references/mcp-contract.md](references/mcp-contract.md) before the first live call or when interpreting a response whose live schema differs from this Skill.

Scope Creep

Low
Category
Excessive Agency
Confidence
75% 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 · references/mcp-contract.md (reported line 284)May include surrounding context.

md
For `queued` or `running` analyses, the analysis status is authoritative. Coverage values are
provisional and must not be interpreted as source failure or exhaustion; wait for the terminal state.
At a terminal state, an evidence family with no records remains part of reported coverage. Treat it as
a limitation; do not resubmit, broaden scope, or discard successful families solely because it is empty.

New result responses return: