Versioned security assessment

Report ID: SA-5B658256

6/30/2026, 3:32:43 AM

agent-tools security assessment v2

Skill Security Certification Report

Audit History
Audit model: codex Historical report
Skill name
agent-tools
Version
v2
Maintainer
inference-sh-3
Source
github.com/inference-sh-3/skills/tree/main/skills/agent-tools/
Coverage
5 Files scanned · 554 Lines analyzed
Policy version
Unavailable

Highest confirmed finding severity

Medium

2 confirmed security findings require attention.

Installation context

Historical evidence

This report may not describe the currently installable artifact. Open the current Skill page for install guidance.

Open current Skill page

This report does not block or authorize the manifest or ZIP.

Static analysis found many command, network, filesystem, and environment-variable patterns. Review shows most command hits are Markdown CLI examples or allowed `infsh` usage, but the skill does recommend `curl | sh`, stores credentials locally, and can submit user data to hosted third-party AI apps. No prompt injection or confirmed malicious intent was found.

Report position

Historical report

Open audit history before using this report to install.

Audit attestation

Not attestable

The required immutable binding is incomplete.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

5 Files scanned · 554 Lines analyzed

3 items shown for review

Limitations

This report does not claim runtime or sandbox execution and does not prove the absence of side effects.

Evidence chain

Follow the evidence from source binding to the install contract. Available evidence supports verification; it is not a safety guarantee.

  1. Source

    Binding unavailable

  2. Artifact

    Identity incomplete

  3. Audit

    Complete

  4. Install contract

    Open manifest to verify

    Open manifest

Capabilities observed

Observed means this report recorded supporting evidence. Not recorded does not prove that a capability is absent.

Contains scripts

May execute code included with the Skill.

Not recorded by this audit

Network access

May connect to external services.

Observed in 4 evidence locations

Filesystem access

May read or write local files.

Observed in 2 evidence locations

Env variables

May read values from the process environment.

Observed in 3 evidence locations

External commands

May invoke commands or programs outside the Skill.

Observed in 4 evidence locations

Risk findings

Confirmed security concerns are separated from items that still need review.

Confirmed security concerns (2)

RISK-001 Medium
SKILL.md:15-17references/authentication.md:5-6references/authentication.md:41-44references/cli-reference.md:5-6
Network Installer Piped to Shell
The skill recommends running `curl -fsSL https://cli.inference.sh | sh`. This is a real supply-chain risk because a remote script executes immediately in the user environment, even though the file also provides a manual install path with checksum verification.
The pipe-to-shell command is directly present in multiple install instructions. The risk is mitigated by the documented manual install option, so this is not enough to prove malicious intent.
RISK-002 Medium
SKILL.md:36-53SKILL.md:71-79references/running-apps.md:119-129
Hosted App Execution Can Send User Data Externally
The core skill purpose is to run inference.sh hosted apps, including LLMs, search, media generation, and Twitter/X actions. This is legitimate functionality, but prompts and inputs may leave the local environment and reach third-party app providers.
The files clearly describe hosted execution and URL-based outputs. This confirms external data flow, but it matches the stated purpose rather than hidden behavior.
Needs review findings (1)
REVIEW-001 Medium
references/authentication.md:15references/authentication.md:25-33references/cli-reference.md:68-72
Credential Handling Requires User Caution
The documentation instructs users to authenticate and optionally set `INFSH_API_KEY`. This is expected for an API CLI, but users could expose account credentials if they paste keys into shared prompts, logs, or shell history.
The API key references are explicit and legitimate for authentication. No evidence shows exfiltration, but credential exposure remains a meaningful operational risk.

Expert evidence

Immutable subject identity, scanner metadata, dismissed matches, and source-level evidence.

Artifact subject

Marketplace commit
Unavailable
Content hash
Unavailable
Tree hash
Unavailable
Skill path
Unavailable
Audit payload hash
Unavailable

Analysis metadata

Audit model: codex

Analysis state: Complete

Scope is limited to the recorded files, lines, methods, and evidence. No runtime or sandbox execution is claimed.

Static false positives ignored (4)
Low
SKILL.md:4SKILL.md:58-67references/app-discovery.md:5-18references/cli-reference.md:42-46
Command Injection Alerts Are Mostly Documentation Examples
Most static external-command detections are Markdown code blocks or command tables showing `infsh` usage. The skill metadata restricts tool access to `Bash(infsh *)`, and no arbitrary shell command construction was found.
The alerts point to documentation and fixed CLI examples, not dynamic code execution. The allowed-tools declaration further narrows execution to the intended CLI.
Low
SKILL.md:20-28references/app-discovery.md:58-60references/authentication.md:33references/cli-reference.md:11-23
SHA-256 Checksum Mention Is Not Weak Cryptography
Static analysis flagged several ordinary documentation lines as weak cryptography. The only relevant crypto context found is SHA-256 checksum verification for downloaded binaries, which is appropriate for integrity checking.
The reviewed static-hit locations do not show a weak cryptographic implementation. The only substantive cryptographic text describes SHA-256 checksum verification for installation.
Low
SKILL.md:29-31references/cli-reference.md:76-84
Filesystem Writes Are Installation and Completion Paths
The hidden-path detections reference installing a binary under the user local bin path and writing shell completions under the Fish config directory. These are expected CLI setup locations, but users should review paths before running commands.
The file writes are visible setup commands, not hidden persistence code. They still touch user configuration paths, so the finding is low rather than safe.
Low
SKILL.md:120-126references/authentication.md:57-59references/running-apps.md:146-148references/app-discovery.md:110-112
Hardcoded URLs Are Expected Documentation Links
The hardcoded URLs point to inference.sh documentation, installer, distribution, output, and related docs. No evidence found of suspicious or unrelated exfiltration endpoints.
The URLs are visible documentation or platform links matching the skill purpose. I did not find unrelated domains or concealed data transfer instructions.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest

The manifest and lockfile bind install artifacts to cryptographic hashes. This integrity claim is separate from the security assessment.

Audit attestation: not_attestable