agentleFS
farukzahrafarukzahra/agent-skills/skills/dont-be-lazy/SKILL.md

dont-be-lazy

Skill0 starsChanged 2 months agoMITView on GitHub

When the user is lazy or refuses manual steps, the agent must not be lazier — discover API/MCP first, store credentials in ../secrets/ (never commit), then drive the Cursor browser to finish the task. Use when the user says lazy, "you do it", or the agent was about to delegate a step the user could avoid.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Golden rule: If the user is lazy, you must not be lazier.

    line 13
  2. Distinctive:

    Before any manual ask, spend real effort finding a programmatic path:

    line 27
  3. Distinctive:

    If you find a working path, use it without asking the user to repeat steps you can run.

    line 37
  4. Distinctive:

    When automation needs a token, password, or session:

    line 41
  5. Very specific:

    If no official API/MCP/CLI exists, or auth requires a human-only gate once:

    line 59
  6. Distinctive:

    After the user unblocks (e.g.

    line 64
  7. Distinctive:

    Only when all of the above failed or a step is legally/financially bound to the human:

    line 71
  8. Distinctive:

    Ask for the minimum single action (one login, one OTP, one approval).

    line 74
  9. Distinctive:

    Ask the user to run commands you can run in the agent shell

    line 80
  10. Distinctive:

    Ask the user to open a URL you can open in the Cursor browser

    line 81
  11. Distinctive:

    Stop after one failed click — diagnose with snapshot/screenshot, adjust, retry (max ~4 attempts per action.

    line 84
  12. Distinctive:

    Before asking the user to do something:

    line 97

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Don't be lazy — outwork the user

Golden rule: If the user is lazy, you must not be lazier. Do not offload work you can still do with API, MCP, secrets, or the Cursor browser.

Pair with automate-before-manual (try automation first). This skill applies when the user pushes back on manual steps or you were about to send a checklist they will ignore.

When to use

  • User says lazy, "you do it", "don't ask me", "just handle it"
  • You were about to ask the user to click, log in, copy-paste, or run something in their terminal/browser
  • A task looks "manual" but might have an API, MCP server, CLI, or browser path you have not tried yet

Escalation ladder (always in order)

1. Discover official automation

Before any manual ask, spend real effort finding a programmatic path:

SourceAction
MCPGetDynamicTools — search namespaces/tools for the product
ProjectAGENTS.md, docs/, scripts/, Makefile, existing PAT/SSH usage
CLIgh, az, cloud/vendor CLIs documented in the repo
HTTP APIOfficial REST/GraphQL docs; prefer documented endpoints over scraping
WebSearch for official API/MCP/CLI — not forum hacks

If you find a working path, use it without asking the user to repeat steps you can run.

2. Credentials → vault, then retry

When automation needs a token, password, or session:

  1. Ask once for exactly what is missing (scope/role included — see automate-before-manual).
  2. Store in the secrets vault (relative path from project root):

    ../secrets/<service>/<purpose>.txt

    Examples: ../secrets/linkedin/farukz.txt, ../secrets/stripe/api-key.txt.

  3. Never commit secrets — confirm .gitignore covers ../secrets/ (vault lives outside repos).
  4. Retry the API/CLI/MCP path with the stored credential before browser fallback.

See reference/credential-vault.md.

3. Browser fallback — you drive, user only unblocks

If no official API/MCP/CLI exists, or auth requires a human-only gate once:

  1. Open the Cursor browser (cursor-ide-browser MCP): browser_navigate → browser_lock → browser_snapshot.
  2. Do the workflow yourself — navigate, fill forms, submit, download, configure.
  3. Pause only for hard blockers: OAuth consent, 2FA, captcha, payment, legal checkbox the user must own.
  4. After the user unblocks (e.g. completes 2FA), resume automation — do not hand back a multi-step checklist.
  5. Validate outcome (HTTP 200, expected page title, file saved) before reporting done.

See reference/browser-fallback.md.

4. Escalate to the user (last resort)

Only when all of the above failed or a step is legally/financially bound to the human:

  • State what you tried (API, MCP, vault, browser).
  • Ask for the minimum single action (one login, one OTP, one approval).
  • Immediately continue yourself after they respond.

Do not

  • Send a long manual checklist when MCP/API/browser could do 80% of it
  • Ask the user to run commands you can run in the agent shell
  • Ask the user to open a URL you can open in the Cursor browser
  • Store or paste secrets into the repo, skills, commits, or chat logs
  • Claim "manual only" without searching for API/MCP and attempting browser automation
  • Stop after one failed click — diagnose with snapshot/screenshot, adjust, retry (max ~4 attempts per action; then report blocker)

Relationship to other skills

SkillRole
automate-before-manualDefault: try PAT/SSH/API before any manual ask
dont-be-lazyStronger: user won't do manual work — you finish via vault + browser
validate-before-shareAfter browser/API work, verify URLs and outcomes before telling the user
find-jobExample: LinkedIn login manual once; agent drives search and Easy Apply

Quick checklist (agent self-audit)

Before asking the user to do something:

  • Searched MCP tools for this product?
  • Checked ../secrets/ for an existing credential?
  • Tried official API/CLI?
  • If still blocked: opened Cursor browser and attempted the flow?
  • Escalation is a single unblock, not a tutorial?

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.