assess
Skill1 starsChanged 7 days agoApache-2.0View on GitHub
Datapace Cloud Self-Assessment of managed databases on AWS, Azure and Google Cloud, run from Claude Code, Codex, Cursor or Gemini CLI. Checks fit, drafts the read-only permissions, runs the Apache-2.0 datapace-collector from PyPI (https://pypi.org/project/datapace-collector/, source in its source distribution) on this machine, then hands off to the browser. Use when the user wants to find cost or waste in their cloud databases, or types /assess.
- Reads credentials
16 other projects use a version of this.
Adopted by
Most recently taken up by
- akshay-diwadkar/first-four-hourstheir own version, first seen Nov 16, 2026
- felipeam86/pipe-skillstheir own version, first seen Nov 15, 2026
- baney75/myelinatetheir own version, first seen Oct 20, 2026
- awslabs/service-onboarding-accelerator-on-awstheir own version, first seen Oct 20, 2026
- rivendale/ai-redteamtheir own version, first seen Oct 17, 2026
- mehrad-dm/mastermindtheir own version, first seen Oct 13, 2026
Used with
Projects with it also use blueprint, verify and 1 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
WordPressSkillblueprint
Use when the deliverable is WordPress Playground Blueprint JSON or a Blueprint bundle, including creating, editing, reviewing, validating schema keys, choosing steps/resources, and debugging Blueprint files. For only running or sharing a Playground environment, use wp-playground.
Also in 3 projects that use this
Microsoft AzureSkillverify
End-to-end app testing pipeline for browser-based web apps on cloud browsers via Azure Playwright Workspaces: verify the live app, author tests for working flows, heal failing tests, then re-run with video and produce one consolidated report. Three phases (app verification, test authoring, test healing) plus a scoped final sanity run; each phase emits a sub-report. Use when asked to "verify my app", "make sure my app works and has good test coverage", "do a full QA pass", "add tests where I'm missing coverage", "fix failing tests and add new ones", "smoke test before deploy", "QA my deployment", or any combined verify-and-test ask. Requires Azure CLI (`az login`) and a Playwright Workspace resource ID — bootstraps configs on first invocation. Browser-based web apps with Playwright only — not for unit/API tests, native mobile/desktop, load/perf, or full accessibility audits.
Also in 4 projects that use this
MicrosoftSkillhandoff
Compact the current conversation into a handoff document for another agent to pick up.
Also in 3 projects that use this
Other versions16 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 16
- bjcoombs/ai-native-toolkit★ 32same name
- mehrad-dm/mastermind★ 23same name
- felipeam86/pipe-skills★ 18same name
- saleh-alhaddad/itqan-engineering★ 10same name
- Ic3b3rg/learn-skills★ 9same name
- benyoskovitz/claude-code-workflow★ 8same name
- awslabs/service-onboarding-accelerator-on-aws★ 7same name
- baney75/myelinate★ 6same name
- jhlee0409/omni-harness-kit★ 2same name
- jvalin17/agent-toolkit★ 2same name
6 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 16 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
If you have no shell on the user's machine (for example claude.ai chat), say the skill needs a local coding agent, give the install lines …
- Very specific: line 45
Check uv --version works. If it is missing, point to https://docs.astral.sh/uv/getting-started/installation/ and let the user install it.
- Very specific: line 49
AWS: use only the profile names the user gives you, mapped to their 12-digit account ids.
- Distinctive: line 58
Write them to a file only if the user asks.
- Very specific: line 67
3, preflight failed, nothing written: read the preflight failed: line and tell the user exactly what to fix: the sign-in, the mapping, or …
- Distinctive: line 91
Stop there. Return only if the user says the upload was refused: over 64 MiB means a rerun with a smaller scope.
“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
assess
You guide the user through the Datapace Cloud Self-Assessment. Everything you do is read-only and local: the collector runs on this machine with the user's own access, and the bundle stays here until the user uploads it in the browser. Credentials never leave this machine.
You are the guide: you ask, draft, run and explain. Sign-up, terms, upload and the report happen in the browser, by the user.
Your agent label is the one that names you: claude-code, codex, cursor or gemini-cli, else other. It goes in the hand-off link and in feedback.
If you have no shell on the user's machine (for example claude.ai chat), say the skill needs a local coding agent, give the install lines from the repository README, and stop.
1. Consent line
Say this first, before anything else, in one sentence:
This skill sends Datapace anonymous reports on how this run goes, good or bad (the step, a category, your agent, the skill version, optionally your cloud provider and database engines, the collector version and which data inputs it collected, and a short message; never account, subscription or project ids, ARNs, hostnames, paths or names); say no to me at any time, or set
DATAPACE_SKILL_NO_FEEDBACK=1, and nothing more is sent.
Done when the sentence is said. A no, then or later, holds for the rest of the session.
2. Fit check
Ask which cloud providers, unless the user said. Once a provider is named, offer discovery: you find out on your own, read-only, what the collector can get there (sign-in, scope, databases, billing access, missing roles and APIs), following references/discovery.md. Run it on a yes; it ends on a table and a proposed collector command for the user to approve. Then ask, together, only what discovery left open:
- Which database engines?
- Which accounts, subscriptions or projects, and which regions?
- Billing access: on AWS, a payer profile with Cost Explorer; on Azure, Cost Management Reader; on Google Cloud, a Cloud Billing export to BigQuery?
Answer from references/coverage.md: say which of their engines the assessment covers, partly covers or does not cover.
- Covered (at least one
availablerow matches): continue with the covered providers. If they also run something not covered, send onenot_coveredreport for it. - Not covered (nothing they run is
available): send onenot_coveredreport with their provider and engines (see Feedback), tell them coverage follows these requests and that datapace.ai/assessment lists what is coming, and stop.
Done when you know, from discovery or the answers, for each covered provider: the engines; the AWS accounts with a profile name for each, the Azure subscription ids or the Google Cloud project ids; the regions; and the billing access.
3. Prerequisites
Check uv --version works. If it is missing, point to https://docs.astral.sh/uv/getting-started/installation/ and let the user install it; uv brings its own Python. A user who cannot install anything can run the collector in Azure Cloud Shell or Google Cloud Shell (see references/collector.md).
Sign-in stays with the user, per provider:
- AWS: use only the profile names the user gives you, mapped to their 12-digit account ids. Leave
~/.awsuntouched: the collector opens the named profiles itself and checks each one is the account it is mapped to. - Azure: the user is signed in with
az login, to the tenant of each subscription. - Google Cloud: the user has run
gcloud auth application-default login.
Draft the read-only permissions for the user to hand to their security team:
- AWS:
references/iam-policy.jsonas is: nine read-only actions. - Azure and Google Cloud: the role list and where each role applies, from
references/collector.md.
Write them to a file only if the user asks.
Done when uv and each provider's sign-in are confirmed and the user has the permissions.
4. Run
For each covered provider, take the command discovery proposed, or assemble it from references/collector.md, with the pinned version and its own --out folder. Show it, then run it through your normal permission prompt so the user approves it. The collector reports on stderr; read the whole output. Act on the exit code:
- 2, usage: fix the flag the error names and rerun.
- 3, preflight failed, nothing written: read the
preflight failed:line and tell the user exactly what to fix: the sign-in, the mapping, or which permission to grant where. Wait for them, rerun. - 0, bundle written: go on below. Each
deniedline is a grant that would make the bundle more complete: list them as "grant<action or role>on<account, subscription or project>" and offer a rerun once granted. A Google Cloud API that is not enabled is the same: name the API and the project. Whengcloudlists instances in that project anyway, the cause is the application-default quota project: tell the user to rungcloud auth application-default set-quota-project <that project>and rerun. - 1 or 130: report the error line; send a
blockedreport if the user cannot get past it.
On exit 0, run the bundle check in references/bundle-check.md: scripts/check_bundle.py with discovery's counts gives a verdict, good, good_with_gaps or bad. On bad, work the fix loop there: at most two automatic reruns, each finding mapped to one move (fix the flags yourself, give the user a command, draft a grant, or explain and wait). Then report:
- the verdict, in plain words, and the one bundle to upload (
datapace-bundle-<id>.tgzin the--outfolder), its size, and that earlier bundles there are superseded; - each gap, with its cause and whether a later rerun helps;
- that
aliases.jsonbeside the bundle maps the aliases back to real names: the user keeps it and never uploads it.
You read the collector output and the check's JSON only; aliases.json and the rows in the bundle stay unopened.
Done when each covered provider has a good bundle, or a good_with_gaps one the user agreed to upload, or a bad one after the fix loop with the user told what is left.
5. Hand-off
Give the user:
- the link
https://assess.datapace.ai/sign-up?utm_source=<agent label>&utm_medium=agent&utm_campaign=assessment-skill; - the bundle paths;
- what comes next: sign up with a work email, accept the terms, create an assessment, drop the files on the upload step (one assessment takes a bundle per provider and merges them into one report); the analysis takes minutes.
Internal testing only: when printenv DATAPACE_ASSESS_BASE_URL is set, it replaces https://assess.datapace.ai in the link; DATAPACE_FEEDBACK_URL likewise replaces the feedback endpoint in scripts/feedback.sh.
Stop there. Return only if the user says the upload was refused: over 64 MiB means a rerun with a smaller scope; anything else, read them the app's message and help from it.
Feedback
The consent line covers every report: send without asking, unless the user said no in this session.
- Friction, as soon as it happens: the user hits something they cannot get past or finds a step confusing, a collector warning contradicts what you observed, the bundle check finds a gap, a rerun was needed, or the user has an idea. One report per problem.
- Outcome, once per run: at the hand-off, or when the run stops short of it. The bundle verdict, the reruns and fixes tried, and discovery against what was collected, in counts.
Send each with scripts/feedback.sh, passing the last bundle's path once there is one, following references/feedback.md for the body and the scrubbing rules; the script does nothing when DATAPACE_SKILL_NO_FEEDBACK=1 is set. After each send, tell the user in one short line what was sent.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

