Back to skill

Security audit

molecular-docking

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent for molecular docking, but it delegates authenticated uploads and API destinations to mutable remote documentation without a local allowlist or integrity check.

Install only if you trust SciMiner's documentation hosting and credential controls. Before use, confirm uploaded receptor and ligand files are allowed to leave your environment, and prefer short-lived, narrowly scoped SciMiner credentials with destination/origin checks outside the skill when 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:185
Finding

Unpinned Remote Documentation Controls Authenticated Requests

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 185-196
Vulnerability Type: Unpinned supply-chain instructions affecting authenticated API requests
Risk Level: Medium

Vulnerable Code Snippet

markdown
### Prerequisite

Use the runtime `SCIMINER_API_KEY` as the `X-Auth-Token`. Do not request, derive, print, persist, or search for the key. If it is absent, stop and report that the SciMiner gateway did not inject the credential.

### Authoritative documentation

Before every invocation, read the selected Markdown file under `https://sciminer.tech/tool_api_files/`:

- `Gnina` -> `Gnina_api_doc.md`
- `AutoDock Vina` -> `AutoDock Vina_api_doc.md`
- `PackDock` -> `PackDock_api_doc.md`
- `SurfDock` -> `SurfDock_api_doc.md`
- `DiffDock` -> `DiffDock_api_doc.md`
- `fpocket` -> `fpocket_api_doc.md`

Treat the current tool doc as the sole source of truth for base URL, endpoint, content type, authentication header, provider/tool names, method, parameters, enum values, upload fields, request encoding, and example submission flow.

The resulting upload and submission behavior is specified at lines 203-208:

markdown
### Submission sequence

1. Select the engine and matching doc; run `fpocket` first when no defensible site exists.
2. Upload each required file exactly as documented and replace local paths with returned `file_id` values.
3. Submit each independent run with its recorded parameters and provenance.
4. Poll and collect all results into the pose pool. For tasks without a fixed ETA, stop polling after 1800 seconds and return the `history_url` for later inspection.

Technical Analysis

The Skill makes mutable remote Markdown the sole authority for security-sensitive request properties, including the base URL, endpoint, authentication header, request encoding, and file-upload flow. No locally pinned endpoint, exact destination allowlist, immutable documentation version, int ...[truncated 2714 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin the approved SciMiner API origin and endpoint patterns in the locally audited Skill or another trusted local configuration.
  2. Enforce an exact HTTPS allowlist before attaching SCIMINER_API_KEY or uploading files. Reject changes in scheme, hostname, port, or unapproved path.
  3. Prevent authenticated requests from following redirects to another origin. Strip credentials on every redirect and fail closed on cross-origin redirects.
  4. Pin each remote documentation file to an immutable version and verify a trusted cryptographic digest or digital signature before using it.
  5. Do not allow remote documentation to redefine the credential header, authentication destination, or permitted upload origins.
  6. Validate documented parameters against a local schema containing approved methods, fields, content types, and destination constraints.
  7. Require explicit user confirmation before uploading receptor or ligand files that may contain confidential or proprietary research data.
  8. Use narrowly scoped, short-lived credentials restricted to the required docking operations, with quotas and server-side destination controls where possible.
  9. Log the validated destination origin, documentation version, and integrity identifier without recording credentials or sensitive molecular contents.
  10. On any integrity, origin, or schema-validation failure, stop without transmitting the token or uploading files.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Static analysis

No suspicious patterns detected.