Complete the delivery
Skill ralfyishere/piensalo/micro-skills/complete-the-delivery
PIÉNSALO is the open artificial cortex for AI—reduce context tokens, verify response quality, expand when needed, and safely fall back.
npx -y skills add ralfyishere/piensalo --skill complete-the-deliveryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 25 days oldThe repository was created 25 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 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
Enumerate every deliverable the request actually asked for and map each one to the work produced, so nothing asked-for is silently dropped, stubbed, or rounded up to done. Activate when a request contains multiple parts — an enumerated list, several 'and'/'also' clauses, or a spec with more than one required artifact — or before delivering a draft against such a request.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.1 KB, 562 tokens by cl100k_base, as published. Nobody here has run it
complete-the-delivery
Trigger (observable): The request contains multiple parts — an enumerated list, several 'and'/'also' clauses, or a spec with more than one required artifact — or a draft is about to be delivered against such a request.
When NOT to activate: Single-deliverable requests with no sub-parts; exploratory conversation where nothing is being delivered; requests explicitly scoped down by the user mid-task (the latest instruction wins).
Procedure
- Extract every requested deliverable from the request verbatim into a checklist — including parts embedded mid-sentence ('...and update the index'), format directives, and counts ('3 examples each').
- Map each checklist item to the specific place in the produced work that satisfies it (file, section, line).
- For any unmapped item: either complete it now, or declare it explicitly undone with the reason — never let a summary claim ('done', 'all set') span it.
- Partition the final claim: delivered-and-verified / delivered-unverified / not delivered. 'Unverified' is its own category, never merged into either neighbor.
- Re-read the original request once, end to end, after the mapping — requests drift out of working memory during long tasks.
- Blocked-capability rule: if delivering requires a capability unavailable in this session (file-write permission, a tool, network), do not stop at a permission request — deliver the complete artifact inline in the answer and state the substitution in one line.
Required output
A deliverable checklist with a mapping (item → location) and an explicit status per item; the completion claim quantified ('4 of 5 delivered; #5 undone because X'). Place the checklist BEFORE the deliverable; never append anything after a task-mandated final line.
Verification
- Every deliverable extractable from the request appears in the checklist with a mapped location or an explicit undone-with-reason status.
- No 'done'/'complete' claim spans an item marked unverified or undone.
Known risk: Checklist theater on trivial two-part requests. Mitigation: below 3 parts, the mapping is one line, not a table.
Max intended cost: ≤250 added output tokens; one re-read of the request.
Evidence status: EXPERIMENTALLY_TESTED — micro-skill repair content lifted a weak model on procedural tasks in controlled runs; the effect is task-concentrated, and automatic router selection of micro-skills is a separate, unproven layer.
Lineage: Grounded in an evidence-backed completion-integrity repair pattern (completion claims scoped to what was actually delivered and verified) and a documented reasoning-failure mode: confidence presented at a level the evidence does not support.