Oathe Security Badge

oathe / scan / reports / safe

Is sherlock-project/sherlock safe?

Yes. sherlock-project/sherlock is safe to install. Oathe's behavioral security audit gave the sherlock skill by sherlock-project a trust score of 88/100 with 8 findings (1 high). Report updated

https://github.com/sherlock-project/sherlock

88
SAFE

Is sherlock-project/sherlock safe to install?

Sherlock is a well-known, legitimate open-source OSINT Python tool with no prompt injection, no credential-stealing code, clean clone behavior, and intact canary files. The primary concerns are behavioral: the repository has no SKILL.md (not designed as an LLM skill), makes mass concurrent HTTP requests to 400+ social media sites by design, and its core purpose — tracking individuals' online presence across platforms — carries meaningful stalking and privacy violation potential when placed in an autonomous agentic context. No technical malice was detected, but the tool's use case requires explicit authorization and careful policy controls before deployment.

What security issues were found in sherlock-project/sherlock?

Category Scores

Prompt Injection 100/100 · 30%
Data Exfiltration 72/100 · 25%
Code Execution 85/100 · 20%
Clone Behavior 95/100 · 10%
Canary Integrity 100/100 · 10%
Behavioral Reasoning 60/100 · 5%

Findings (8)

MEDIUM No SKILL.md — Not Designed as LLM Skill -20 ▶

The repository contains no SKILL.md file. This is a standalone Python CLI tool not packaged or intended as an LLM agent plugin. Installing it as a skill requires additional wrapping that is not provided, and the tool's behavior is not constrained to a declared skill contract.

Evidence: SKILL.md content is empty; tool entry point is sherlock CLI or python -m sherlock_project

HIGH OSINT Username Tracking Tool with Stalking/Surveillance Potential -20 ▶

Sherlock's primary purpose is discovering a person's presence across 400+ social networks. In an agentic context, it could be weaponized for unauthorized individual surveillance, profile compilation for social engineering, or stalking. The tool includes NSFW sites and covers a broad range of platforms.

Evidence: sherlock.py docstring: 'Find Usernames Across Social Networks'; data.json entries include 400+ platforms; test_ux.py references NSFW site filtering

MEDIUM Mass Concurrent Outbound HTTP Requests to 400+ Sites -18 ▶

Sherlock's core function sends parallel HTTP requests (up to 20 concurrent workers) to hundreds of external social media platforms. While by design, this represents a large outbound traffic footprint that could cause rate abuse, ToS violations, or attribution when invoked by an autonomous agent.

Evidence: sherlock.py: SherlockFuturesSession(max_workers=max_workers); data.json contains 400+ site entries

MEDIUM Remote JSON Data Loading Accepts Arbitrary URLs -10 ▶

The --json parameter accepts any URL or a GitHub pull request number, dynamically downloading site manifests. An attacker who controls the data source could supply a site list containing malicious probe endpoints, causing the agent to make requests to attacker-controlled servers.

Evidence: sherlock.py main(): json_file_location = args.json_file; PR number support constructs raw.githubusercontent.com URL from PR JSON

LOW Executable Apify Actor Shell Script Included -8 ▶

The .actor/actor.sh shell script (chmod 775) is included for Apify cloud actor platform integration. While a legitimate integration pattern, it is executable shell code outside the Python package proper.

Evidence: .actor/actor.sh inode=270608 mode=0100775; .actor/Dockerfile also present

LOW GitHub API Version Check on Every Invocation -5 ▶

Each invocation contacts the GitHub API to check for a newer version. This telemetry is minor but represents an unconditional outbound call that could reveal agent activity patterns.

Evidence: sherlock.py main(): requests.get(forge_api_latest_release, timeout=10).text

INFO Clean Clone — Only Expected GitHub Connections 0 ▶

The git clone made only expected HTTPS connections to GitHub (140.82.121.3:443). No unexpected network activity, process spawning, or filesystem writes outside the skill directory were observed. No git hooks or submodules were present.

Evidence: Network log: only 140.82.121.3:443 (GitHub) and 185.125.190.48:443 (Canonical Ubuntu) observed during clone window

INFO All Canary Honeypot Files Intact 0 ▶

None of the honeypot credential files (.env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, .config/gcloud/application_default_credentials.json) were accessed by any code originating from the cloned repository. The early accesses visible in the audit log are attributable to the oathe monitoring infrastructure itself.

Evidence: Canary integrity report: ✅ All canary files intact; early file accesses at audit timestamp 1794748607 (pre-clone) are from the audit bootstrap process

Should I install sherlock-project/sherlock?

Oathe's verdict for sherlock-project/sherlock is SAFE with a trust score of 88/100. Recommendation: Review before install.