awesome-claude-code review Turning a directory into a useful shortlist
A pinned inventory audit of 263 records, 252 active links and six public destinations, covering categories, check dates, legacy entries and licensing boundaries.
AgentIndex · Published 2026-11-04 · Updated 2026-11-04
awesome-claude-code helps with the discovery stage of choosing tools for Claude Code. If a team wants better session memory, a testing workflow or a plugin, a categorized collection gives it somewhere to start. The adoption decision still belongs to the team, using the individual project's implementation, documentation and supported environment.
Our recommendation is to use this collection to produce a small shortlist. Its entries do not form a single software distribution with shared compatibility, licensing or acceptance tests. We audited the directory itself and visited six sampled destinations. We did not install any listed skill, plugin, client or MCP server.
What we checked
On November 4, 2026, we fixed the upstream revision at a4600018e7cb6282cceae22cd9a354c6cf548a98. The audit reads the README, resource CSV, contribution guide and license. It checks their SHA-256 fingerprints before calculating results, so a future revision cannot silently inherit this snapshot's numbers. Pinned resource table
| Check | Observed result |
|---|---|
| All resource records | 263 |
| Records with Active set to TRUE | 252, across 24 categories |
| Record identifiers | All 263 unique |
| Active resource URLs | All 252 unique |
| Active URLs represented in README | All present |
| Empty Last Checked values | 46 records, counting the whole table |
| Six sampled public destinations | All returned HTTP 200 during this run |
The sampling rule takes the first two active entries from each of three categories: official Anthropic resources, memory and context persistence, and testing. Keeping that rule visible matters. A reader can reproduce the same sample rather than assume we selected whichever six links happened to work.
Our verification fixture contains the audit, file fingerprints, sampled addresses and output. It does not redistribute the complete upstream collection. The inventory checks require a separate checkout of the named revision; the six-link check additionally requires network access.
Start with a problem, then a category
Categories become useful when they express the problem being investigated. If developers repeatedly lose project conventions between sessions, the memory category offers a manageable starting point. If changes lack a repeatable verification step, the testing category is more relevant. Separating official resources also makes it easier to understand native product features before adding another dependency. Directory organization
However, two entries in one category may have very different operating models. One might be a document, another a long-running service, and another a script invoked by a hook. Before comparing their promises, write down where each component is installed, when it runs, which files it can touch and how it is removed. Those answers make the choice more concrete.
For an initial comparison, choose two candidates and define one small task. A memory trial could save a deliberately invented project convention and ask for it in a fresh session. Record setup steps, observable output and cleanup. Such a trial would be the adopting team's own evaluation: our directory audit does not establish how those projects perform.
Interpret maintenance fields carefully
The CSV exposes Active, Last Checked and Stale fields. In this snapshot every Stale value is FALSE, while 46 records have no Last Checked date. An empty date does not establish that the corresponding project is broken. Conversely, a FALSE stale flag does not establish that somebody recently completed its installation and verified its behavior.
Treat those fields as prompts for investigation. Open the candidate's repository and inspect its actual release, installation entry point, supported host versions and relevant issue discussions. If the command has changed, use the candidate project's current instructions when designing the trial. A directory description is too small to carry every migration detail.
The README also explains that the current collection is being rebuilt and that older resources are preserved under README_ALTERNATIVES, with maintained entries migrating back. A project missing from the main table has therefore not necessarily failed a quality review. Readers looking for a familiar entry should check which generation of the list they are viewing. Rebuild explanation
Curation helps narrow the search
The contribution guide describes eligibility based on project age and continued activity, or a star threshold. It also says that recommendations receive best-effort attention, with no guaranteed response or inclusion. Mechanical issue-form validation is distinct from assessing a resource's quality. These details help explain how the collection is assembled. Contribution rules
They do not provide an acceptance test for your environment. A plugin that will access private project files needs a specific version, understood configuration and a trial in an appropriate test project. A learning document needs a different kind of check: whether its examples still describe the product being used. Apply the standard that matches the resource instead of expecting one badge or inclusion decision to answer both questions.
HTTP reachability has an equally narrow meaning. Our six successful responses show that those destinations were reachable during one run. They say nothing about installability, account requirements, compatibility, security properties or future availability. We make no claim about the other active links' network status.
Read the license at both levels
The collection's license at this revision is CC BY-NC-ND 4.0. That applies to the collection itself; linked projects have their own terms. A directory entry does not transfer its license to a destination or establish that the destination permits commercial use. Anyone planning to copy or adapt the collection should read the original license and maintainer's accompanying explanation. Collection license
For day-to-day selection, aim to leave with a one-page shortlist: the problem, two candidates, their versions and original licenses, and the result of a repeatable small task. The directory contributes the discovery step. Evidence from the individual candidates supplies the rest of the decision.
If the immediate goal is to understand agent execution patterns, continue with our GenAI_Agents tutorial review. If the work involves validating tool data, the datamodel-code-generator contract review examines a narrower component with executable input and output checks.