Workflow graduation
Curated agent skills for Cursor — diagnose, triage, commits, workflows.
npx -y skills add TonkaToyXL/cursor-agent-skills --skill workflow-graduationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when a task, fix, research flow, content pipeline, or agent workflow repeats or is likely to repeat. Converts one-off work into durable SOPs, skills, scripts, checklists, memory notes, or backlog items so the agent compounds instead of rediscovering.
SKILL.md
3.7 KB, 779 tokens by cl100k_base, as published. Nobody here has run it
Workflow Graduation
Purpose
Turn repeated or high-value agent work into durable operating assets. This is the “don’t rediscover it next time” skill.
Use it when:
- A workflow has been done twice or will clearly recur.
- A bug/fix/config issue repeats.
- A user says “make this reusable,” “do this properly,” “remember the process,” or “follow your recommendation.”
- A task produced a useful pattern, checklist, prompt, source set, command, or acceptance gate.
- A workflow belongs to a named lane such as content pipelines, config repair, research briefs, or agent/tool setup.
Do not use it for true one-offs with no future value.
Graduation Decision
Classify the work into the smallest durable artifact that will prevent future rediscovery:
| Pattern | Graduate into |
|---|---|
| Stable multi-step workflow | Skill (~/.agents/skills/<name>/SKILL.md) |
| Small command sequence | Script or launcher |
| Repeated checklist | SOP/checklist markdown |
| Durable preference/context | Memory note / Obsidian / Gemini Memory |
| Bug or trap | Mistake note + guardrail |
| Future improvement | Backlog item with trigger + acceptance criteria |
Prefer the least complex artifact that will actually be used.
Required Workflow
-
Capture the trigger
- What happened?
- What phrase/task should trigger this next time?
- Which lane does it belong to?
-
Extract the reusable shape
- Inputs needed
- Steps that worked
- Tools/sources used
- Verification gate
- Output contract
- Known caveats
-
Choose the artifact
- Skill for cross-session behavior
- SOP/checklist for guidance
- Script for deterministic commands
- Memory for durable preferences/facts
- Backlog if not implemented now
-
Install or update it
- Put active skills under
~/.agents/skills/<skill-name>/SKILL.mdor your agent skills directory. - Put workspace SOPs/backlogs under your agent workspace memory folder.
- Do not edit generated foundation blocks directly unless the user explicitly asks.
- Put active skills under
-
Verify
- YAML frontmatter parses.
- Trigger description is clear enough for future skill selection.
- Artifact path exists.
- If a script was added, run a dry-run/help/syntax check.
- If a workflow was added, include an acceptance gate.
-
Report briefly
- What was made
- Where it lives
- How it will trigger next time
- Any remaining backlog
Backlog Item Template
### <title>
- **Lane:** <agent / content / client / personal ops>
- **Trigger:** <when this should apply>
- **Problem:** <what keeps getting rediscovered or repeated>
- **Durable asset:** <skill / script / SOP / memory / guardrail>
- **Acceptance gate:** <how future agent verifies it worked>
- **Status:** queued | in-progress | shipped | retired
Quality Bar
A graduated workflow is only “done” if a future session can execute it without asking the user to re-explain the context.
Good assets are:
- Short enough to be read quickly.
- Specific enough to prevent mistakes.
- Source-backed when claims matter.
- Paired with a verification gate.
- Updated when reality changes.
Anti-Patterns
Avoid:
- Huge generic skills that trigger on everything.
- Saving secrets or private tokens.
- Backlogs with no trigger or acceptance gate.
- Writing “remember to…” without creating a durable file/tool.
- Making a skill when a 10-line checklist would do.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most automation workflows skills give in 779 tokens
Counted across 745 of the 1,008 authors here whose files we hold, read 2026-08-07
- Write conventional commit messagesin 36 of 745, across 35 files
- Delete branches after mergein 30 of 745, across 21 files
- Make atomic commitsin 25 of 745, across 15 files
- Write minimal code to pass testsin 22 of 745, across 10 files
- Re-snapshot after navigation or DOM changesin 21 of 745, across 13 files
- Use try-catch for error handlingin 20 of 745, across 8 files
- Run tests before committingin 20 of 745, across 12 files
- Write tests before implementationin 20 of 745, across 8 files
- Configure branch protection rulesin 19 of 745, across 5 files
- Explain the why in commit messagesin 19 of 745, across 9 files
- Refactor code while tests remain greenin 19 of 745, across 6 files
- Interact with elements using refsin 19 of 745, across 11 files
Said here and by no other author read
- Convert recurring work into durable artifacts
- Classify work into the smallest useful artifact
- Prefer the least complex artifact that will be used
- Capture the trigger and workflow lane
- Extract inputs, steps, tools, and verification gates
- Install artifacts in the correct directory
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.