Prepare clear continuation notes for another developer or AI agent picking up the work.
SkillScore100Maintained
Automatic
◆Knowledge Management pack
Last verified 8/3/2026
What it changes
Passed 22 of 22 test cases with the skill, and 17 of 22 without it.
Without skill77%
With skill100%
Checks it passed that the model alone missed
The output separates mandatory immediate next steps from non-critical items.
The output highlights static tax rates as an assumption requiring validation.
The output lists 'services/stripe.ts' and 'services/email.ts' as touched files.
and 3 more in the full report
5 fixed17 passed either way0 still failing0 broke
Pass-rate lift
+23 pp
Tokens
+39%
Agent turns
0%
Response time
-11%
Model gemini-3.6-flashMeasured 8/3/2026v122 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
I have completely finished the bug fix for issue #402 in 'utils/date.ts'. All unit tests pass, regression suite passed, code review comments were addressed, and changes were merged to main. Provide a task completion response.
What was checked, without and with the skill
Checked
Without skill
With skill
The output confirms final task completion without handoff continuation sections.
Missed
Passed
case-20 · gemini-3.6-flash · 8/3/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md58 linessingle file
Handoff Notes Skill
Purpose
Give the next person/agent everything they need to continue the work efficiently: what's done, what's in progress, what's left, and anything they need to know before continuing.
When to Use
A task is incomplete at the end of a session and work will resume later.
Explicitly asked for handoff/continuation notes.
A task is complete but has follow-ups, caveats, or context the next person should know.
When Not to Use
The task is fully complete with no follow-ups, and the Global Agent Output and Reporting Skill's summary already covers everything needed.
Required Inputs
Current state of the work: what's done, what's in progress, what's not started.
Any blockers, open questions, or assumptions made.
Relevant file paths and context needed to resume.
Assumptions
The reader has access to the repo but not to this conversation's full context — notes must be self-contained.
The reader may be a different AI agent or a human developer — write for both.
Step-by-Step Workflow
Summarize the goal: what was the task trying to accomplish?
State current status: done / in progress / not started, in concrete terms (not "mostly done" — say what specifically works and what doesn't).
List files touched so far and their role (created/modified, what each contains/does).
Describe what's left, as concrete next steps, in suggested order.
Note blockers/open questions: anything that needs a decision or external input before continuing (e.g. "needs design confirmation on X", "waiting on API endpoint Y to exist").
Note assumptions made that the next person should be aware of/validate.
Note any known issues: things that don't work yet, edge cases not handled, validation not run.
Point to relevant context: related plan/decisions, existing patterns followed, anything non-obvious about why something was done a certain way.
Rules and Best Practices
Be specific: "Implemented the form UI in components/SignupForm.tsx; not yet wired to the signup API" is useful, "mostly done with signup" is not.
Order next steps logically (e.g. respecting build order from Plan the Feature Skill).
Separate "must do next" from "nice to have later."
Include enough context that someone unfamiliar with this specific conversation can pick up without re-deriving decisions already made.
Keep it concise — a focused list beats a long narrative.