Skip to content
TokRepo project interface
Partial

Public pages and primary sources were checked, but the core workflow was not independently completed or audited.

TokRepo

Its interesting idea is not another prompt directory, but a single agent workflow for discovery, risk disclosure, install planning, dry runs, rollback, and machine-readable evidence. Content hashes can show that bytes have not changed, but cannot replace publisher identity, licensing, or human review.

Original author
Regulus K.K.
Last verified
2026-07-18

01 / ANALYSIS

Product analysis

Verification profile
Maturity
Live
Confidence
Partial
Verified
2026-07-18
Commercial relation
Not sponsored

Problem solved

Agent skills, system prompts, MCP configurations, and workflows are scattered across repositories, documentation, and chat history. Before copying them, teams lack a consistent way to assess provenance, permissions, dependencies, target tools, write scope, rollback, and reusable install records.

Why it was selected

The public catalog is browsable without an account, while the official site also exposes CLI, MCP, API, trust manifests, policy packs, and evaluation evidence. Its default policy puts verification and install planning before writes, then requires confirmation or staging for code execution, network access, secrets, or global configuration. That sequence is more auditable than handing a one-line install command directly to a privileged agent.

What is genuinely novel

Registries, package managers, and hashes are not new. The more distinctive move is reshaping them into agent-callable contracts: search, detail, verification, permission envelopes, install plans, dry runs, staging, lifecycle manifests, and rollback are exposed through CLI, MCP, and machine-readable endpoints.

Best for

Developers who need reusable skills, prompts, or MCP configurations across AI coding tools and are willing to verify provenance and write scope before activation

Editorial evaluation

Problem insightstrong
Originalitystrong
Executionmedium
Transferabilitystrong

KEEP DISCOVERING

A few more products worth noticing

A small set selected by product shape and profile similarity, with independent verification records.

Browse all products

02 / WHAT STANDS OUT

Three product decisions worth noticing

01

Separate discovery, verification, and activation

Search should narrow candidates, verification should expose provenance and permissions, and only activation should gain write access. Each layer should be independently stoppable and leave evidence.

02

Integrity is not identity or licensing

Content hashes are useful for detecting changes, but publisher verification, source repository, signatures, and licensing need separate fields and must not be collapsed into one trust score.

03

Machine manifests need continuous testing too

Agents treat JSON contracts as operational truth. Version drift, polluted arrays, and schema gaps should block release rather than being tolerated because the human-facing page still works.

Product insight worth carrying forward

Separate finding an asset from allowing it into an agent context: return candidates and evidence first, produce a reviewable install plan second, and write only after policy gates pass.

03 / DO NOT COPY

What not to copy

Do not hand a one-line install command from search directly to an agent with global write access. Inspect raw files, publisher, and license; then review permissions, pin a version, and use dry-run or staging.
Do not describe default telemetry only as 'anonymous aggregate' data. Persistent identifiers, transmitted fields, purpose, retention, deletion, and opt-out need explicit privacy disclosure.

04 / LIMITS & RISKS

Limits and risks

Sample public API records can simultaneously show an 'established' author trust level, an unverified publisher, unreviewed status, and hash-only signatures. A hash proves integrity, not publisher identity, licensing, or safety. No public terms page was found at verification, and the asset-metadata schema has no license field. The latest npm CLI was 3.23.5 while the trust manifest still named 3.23.4 and the public GitHub main branch package.json named 3.13.1; the latest CLI package exposed neither npm provenance nor a gitHead, so its exact source could not be reproduced from the public repository. Telemetry can be disabled, but package code creates a persistent machine identifier and caller hash by default and sends events, asset IDs, and limited metadata; the current privacy page does not enumerate those identifiers, providers, exact retention, or deletion path. A machine manifest also mixes repair-eval URLs into target, install-mode, and recommended-file arrays, showing that the machine-readable contracts themselves still need version validation.

Compared with alternatives

Compared with copying text from GitHub or a directory, TokRepo adds structured risk, target adapters, dry runs, and rollback. Compared with npm or PyPI, it can distribute instructional assets but lacks an equally mature publisher-identity and licensing chain. Compared with closed official plugin marketplaces, it is broader and more open, while leaving more verification responsibility to the user.

05 / EVIDENCE

Visual evidence

Current TokRepo homepage showing asset search, agent connection, and the public usage funnel
A 1265×712 capture of TokRepo's official English homepage on July 18, 2026. It verifies the public interface and entry points, not the displayed usage numbers, install safety, or asset quality independently.

06 / SOURCE

Source and verification

Original author
Regulus K.K.
Country / region
Japan / global
Maturity
Live
Confidence
Partial
Discovered
2026-07-18
Last verified
2026-07-18

What this review checked

We verified the live site, About and Help pages, April 2026 privacy policy, public API, trust manifest, default policy, install-safety eval, GitHub repositories, npm metadata and package code, the Uneed launch page, and a current screenshot of the official interface. We did not register, generate an API key, install or run the CLI or MCP server, call the hosted MCP endpoint, write any asset locally, publish content, or provide private files, so search quality, install success, and rollback reliability remain untested independently.

Latest change

Initial profile created from the live site, privacy policy, public API, machine manifests, policy and eval evidence, GitHub, npm packages, Uneed launch page, and a real official-interface capture; no install, sign-in, or publishing occurred, and hash, identity, licensing, version, telemetry, and manifest-quality boundaries are recorded.

Ready to explore further?

Open the product, or return to the original source to verify the details.

ONE-TAP FEEDBACK

Did this profile help you judge the product?

One tap. No writing required.