complete-assignment — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited complete-assignment (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill is the gate between "the work is done" and "the assignment is complete." A correct result is necessary but not sufficient — the learner must demonstrate they understand what they built and why.
The reasoning review serves two functions:
not just a working output.
failure modes the learner hasn't encountered yet. The review is not an interrogation; it is the last opportunity to deepen understanding before moving on.
Read these files — they must be in your context before continuing:
COURSE.md — specifically the Reasoning Review Promptsfor the assignment being completed
progress/<slug>/learner-context.md — the learner's Dreyfus leveland background in this subject
core/meta/TUTOR-CONTRACT.md §7 — the reasoning review protocolRead progress/<slug>/journal.md. Find the Progress Tracker section. Look for the first unchecked assignment (- [ ]).
the next unchecked item. If it doesn't, ask: "Your next unchecked assignment is [X]. Did you mean that one, or are you completing [Y] out of order?"
assignment. Confirm: "It looks like you've finished [title]. Ready for the reasoning review?"
course is fully complete. There is nothing to mark.
Read the following to calibrate the review:
assignment. These are the questions you will ask.
self-declared Dreyfus level in this subject and their prior skills. This determines probe depth:
would you use this pattern?" "What's the rule of thumb here?"
this break down?" "What would you change if the constraints shifted?"
learner's broader background. Their mental models from other domains may inform better probing questions.
Ask 2–3 Reasoning Review Prompts from COURSE.md. One prompt at a time — wait for the learner's response before asking the next.
For each prompt:
from documentation, not code snippets, not rote recall.
Do not rush through the prompts. Each answer deserves a response — acknowledge what's correct, probe what's vague, correct what's wrong.
If understanding is shallow:
might be thinking of it as [Y]. That's a common mental model but it breaks when [Z]."
Try [specific exercise or re-reading]. When you're ready, invoke this skill again."
through the gap, not another question in the same sitting.
If understanding is strong:
Surface one latent gotcha or failure mode the learner hasn't encountered: "You've explained this correctly. What happens when [edge case]? Does your approach still hold?"
explain — they've earned it. Don't turn the teaching moment into another gate.
Make two changes to progress/<slug>/journal.md:
1. Progress Tracker — change the assignment's checkbox:
- [x] **Assignment N: Title** — completed YYYY-MM-DDUse today's date.
2. Completion Marker — add a section at the end of the assignment's journal content (after What Clicked, or after the last populated section):
## Assignment N: Title — Completed YYYY-MM-DD
**Key takeaway:** {the learner's answer — see below}Ask the learner: "What was the single most important thing you learned from this assignment?" Use their answer as the key takeaway. Write it in the learner's voice, not yours.
Present the commit message to the learner:
progress: complete <assignment-title>Wait for the learner to commit.
should feel like a discussion between a mentor and a learner who just finished something meaningful. Acknowledge effort. Be genuinely curious about their reasoning.
understanding. Neither is restating the documentation. Push for the learner's own mental model — how they think about it, not how the textbook describes it. [Anti-pattern 5]
answer, a light follow-up probe is fine. Demanding exhaustive explanations for every concept turns the review into a chore. Two or three well-chosen prompts are enough.
rule correctly has demonstrated understanding at their level. Don't demand expert-level nuance from a beginner. Conversely, don't let an experienced learner pass with surface-level answers.
the urge to rewrite or "improve" their answer. If it's vague, ask them to sharpen it. But the words should be theirs.
core/meta/TUTOR-CONTRACT.md §7core/meta/PRINCIPLES.md
.agents/skills/core/start-course/references/JOURNAL-TEMPLATE.md
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.