Back to skill

Security audit

audit-report

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent audit-reporting helper, but it tells agents to run an unbundled Python validator from a dynamically located skills root, which creates a review-worthy local code execution risk.

Install only if you trust the surrounding Groundcrew installation and can verify the .groundcrew/groundcrew-doctor.py script and skills root are controlled by trusted parties. Treat optional third-party SEO connectors as user-authorized data sharing and keep API keys out of reports.

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
SKILL.md:6
Finding

Unvalidated Script Execution from a Dynamically Located Skills Root

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 6
Vulnerability Type: Execution of an unverified local script selected through an unspecified root-discovery process
Risk Level: High

Vulnerable Code Snippet

markdown
Read `references/provider-selection.md` before choosing a source. Detect what is already available, run with it, deliver value, then recommend at most one missing connector. All normalized factual inputs must satisfy the Groundcrew evidence contract. Locate the active skills root and run `<skills-root>/.groundcrew/groundcrew-doctor.py --evidence <record.json>` before using them in a conclusion. Output follows `references/reporting.md`: shared skeleton, SHIP/FIX/BLOCK/UNDECIDED verdict, Measured/User-provided/Estimated labels.

Technical Analysis

The skill instructs the agent to dynamically locate an “active skills root” and execute groundcrew-doctor.py from that location. It does not define a trusted canonical path, verify file ownership or permissions, validate a cryptographic digest, or otherwise establish that the selected script belongs to a trusted installation.

The audited package does not contain .groundcrew/groundcrew-doctor.py, so the executable is outside the reviewed project boundary. Its behavior can therefore differ from the behavior implied by the reviewed skill files. If an attacker can influence skills-root discovery or write to the selected directory, the attacker can supply a malicious script at the expected path.

This is an insecure coding and configuration practice rather than an embedded malicious payload: the vulnerable instruction creates a local code-execution trust boundary without specifying how that boundary must be authenticated.

Attack Path

  1. An attacker gains write access to a directory that can be selected as the active skills root, or influences the agent's root-discovery mechanism.
  2. The attacker creates the expected path: .groundcrew/groundcrew-doctor.py.
  3. T ...[truncated 1407 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace dynamic skills-root discovery with a canonical, administrator-configured absolute path.
  2. Require the script and all parent directories to be owned by a trusted account and reject paths writable by untrusted users or groups.
  3. Resolve symbolic links before execution and verify that the final path remains inside the approved installation directory.
  4. Pin and verify a cryptographic digest or signed manifest for groundcrew-doctor.py before every execution.
  5. Invoke the script with a fixed interpreter and argument array rather than constructing a shell command.
  6. Validate and canonicalize record.json; ensure it is a regular file in an approved directory and cannot inject command-line options.
  7. Run the validator in a sandbox with minimal filesystem, environment, and network access.
  8. Fail closed when the trusted validator is absent or verification fails. Do not search alternate writable roots.
  9. Include the reviewed validator in the skill package, or declare it as a version-pinned and integrity-verified dependency so that the executable code is within the audit boundary.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
Findings (1)

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/connectors.md (reported line 48)May include surrounding context.

md
- **Tier 1.** Google Suggest for expansion (unofficial endpoint; treat as hints, `confidence: low`). A SERP the user pastes or fetches is an observation for that query, locale, device, and date — record all four.
- **DataForSEO.** Paid, per-request. Read `dataforseo.md`; show the cost preflight and get approval for each batch.
- **Ahrefs MCP** (`https://api.ahrefs.com/mcp/mcp`, API key on a plan that includes MCP) and **Semrush MCP** (`https://mcp.semrush.com/v1/mcp`, OAuth or API key). Use only when the user already pays for the tool and has configured the MCP in their client. Label rows `source: "ahrefs"` / `"semrush"`, `confidence` per the vendor's own accuracy notes; these are third-party estimates, not the user's own data.
- **TrustGrowth.** `keywords` (typed opportunities with volume, difficulty, intent, current position, scores) for the verified site. SERP snapshots per keyword are scheduled (TEC-7796).

### `~~link database` — backlinks and authority