Back to skill

Security audit

Ghostfetch

Security checks for vulnerabilities and agentic risk

Overview

This web-fetching skill is coherent, but it should be reviewed because it installs and runs an externally built binary that is not fully contained in the submitted package.

Install only if you are comfortable building and running code from the referenced Ghostfetch repository. Prefer the pinned setup path over the unpinned manual clone, avoid using persisted cookies for sensitive browsing, avoid passing CAPTCHA API keys on shared systems, and treat the read-only/no-secret-leak claims as unverified unless you audit the upstream source yourself.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Warning
Location
SKILL.md:91
Finding
External Source Retrieval, Build, Installation, and Execution<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md:91-96`; `setup.sh:25-55` **Vulnerability Type**: Remote payload retrieval and execution through an unaudited external source **Risk Level**: Medium ### Vulnerable Code `SKILL.md:91-96`: ```bash git clone https://github.com/neothelobster/ghostfetch.git cd ghostfetch go build -o ghostfetch . cp ghostfetch ~/.openclaw/workspace/tools/ ``` `setup.sh:25-55`: ```bash # Clone at pinned commit REPO_DIR="/tmp/ghostfetch-build" rm -rf "$REPO_DIR" echo "Cloning ghostfetch at pinned commit..." git clone "$REPO_URL" "$REPO_DIR" git -C "$REPO_DIR" checkout "$PINNED_COMMIT" # Verify checkout ACTUAL_COMMIT="$(git -C "$REPO_DIR" rev-parse HEAD)" if [ "$ACTUAL_COMMIT" != "$PINNED_COMMIT" ]; then echo "ERROR: Commit verification failed." echo "Expected: $PINNED_COMMIT" echo "Got: $ACTUAL_COMMIT" exit 1 fi echo "Commit verified: $PINNED_COMMIT" # Build echo "Building ghostfetch..." cd "$REPO_DIR" go build -o ghostfetch . # Install mkdir -p "$TOOLS_DIR" cp ghostfetch "$TOOLS_DIR/ghostfetch" chmod +x "$TOOLS_DIR/ghostfetch" # Verify binary works if "$TOOLS_DIR/ghostfetch" --help >/dev/null 2>&1; then ``` ### Technical Analysis The submitted project does not contain the source code or Go dependency manifests used to build the `ghostfetch` executable. Instead, users are instructed to retrieve source from an external GitHub repository, compile it, install it into the OpenClaw tool directory, and execute it. The installation command documented in `SKILL.md` clones the repository's mutable default branch without checking a commit, release signature, or content digest. Consequently, the source compiled by a user can differ from the source that existed when this Skill was reviewed. The included `setup.sh` provides a meaningful mitigation by checking out and verifying commit `6e6876a90470d4bb53e38be32e8f43e67b695b48`. However, the pinned source and its dependency lock data are absent from the s ...[truncated 2206 chars]
Remediation
<![CDATA[ ## Remediation Suggestions 1. Vendor the reviewed Ghostfetch source, `go.mod`, and `go.sum` inside the Skill package so the effective implementation is available during audit. 2. Remove the unpinned `git clone` installation example from `SKILL.md`. Require an immutable commit or signed release for every installation path. 3. Verify the downloaded source or release artifact against a cryptographic SHA-256 digest stored in the reviewed package. 4. Build with locked dependencies and enforce module checksum verification. Consider a vendored dependency tree or an approved internal module proxy. 5. Use reproducible builds and verify the final executable against a published expected digest before installation or execution. 6. Do not automatically execute the newly built binary as a setup verification step. If verification is necessary, perform it only after integrity checks and in a restricted environment. 7. Install the tool into a Skill-specific, non-global location where possible, rather than a shared tools directory available across sessions. 8. Document that search queries, requested URLs, cookies, and optional CAPTCHA credentials are transmitted to external services. Recommend `--no-cookies` by default and avoid passing CAPTCHA API keys through command-line arguments, where they may be exposed in process listings or shell history. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding
The skill exposes shell-capable usage examples and requires a local binary, but it does not declare any explicit tool scope such as permissions or allowed-tools. In an agent setting, that omission weakens policy enforcement and can allow broader-than-intended shell execution or make review of the skill's real capabilities harder.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The documentation states that cookie jars may persist across requests, but it does not clearly warn that retained cookies can carry authentication state, tracking identifiers, or other sensitive session data between fetches. In an agent workflow, this can cause unintended cross-site state reuse, privacy leakage, or requests being made under a prior session context.

External Transmission

Medium
Category
Data Exfiltration
Content
### Scrape structured data
```bash
ghostfetch fetch https://api.example.com/data --json
```

### Find all GitHub links on a page
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Static analysis

No suspicious patterns detected.