Back to skill

Security audit

Build a context pack

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because scanners report that its dependency workflow can send names from local project manifests to public registries, and its helper script was only partly inspected.

Treat this as a review decision before installing. Consider whether you need the dependency lookup workflow and whether its external disclosure of package names is acceptable; the helper script was only partly inspected, so the evidence is incomplete.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (4)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 19)May include surrounding context.

md
Use the bundled Node 22+ helper `scripts/ink.mjs` to fetch sources and preserve a refreshable snapshot:

Self-Modification

High
Category
Rogue Agent
Confidence
85% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

node /scripts/ink.mjs pack --url 'https://llmstxt.org' --url 'https://docs.astro.build/en/getting-started/' --out context-pack.json --markdown context-pack.md

text

Replace sample URLs and filenames with the task's values, safely quoted. The helper uses the public llms.ink API, reads 1–5 sources sequentially, and refuses to overwrite existing files. Default mode is `full`; use `--mode llms` if the user needs indexes only. Split larger requests into explicit batches rather than silently dropping sources. The service has shared per-IP quotas and a 24-hour cache; one source's failure stops the run and is not represented as a successful empty source. No key is needed; an optional `LLMS_INK_KEY` is sent as a Bearer token for its own quota, and an invalid key stops the run with HTTP 401.

Without Node, read each source with the MCP `llms_full_txt` tool (or `llms_txt` for indexes), or fetch `https://llms.ink/full/<url>`, and write the Markdown pack yourself: one section per so

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The deps workflow reads local project manifests and sends dependency names to public registries, beyond the declared public-URL collection workflow.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

The deps command extracts dependency names from local manifests and transmits them to external registries without checking whether they identify private packages or requiring disclosure-specific approval. This can expose internal package names, organizational scopes, and technology choices beyond the advertised public-URL collection workflow. Exposure requires invocation of deps and is limited to queried package identifiers rather than complete manifest contents.

Content

No source excerpt is available for this finding.