15.20Academic University ResearchSupervisorsAvailable

Supervision Feedback and Development

Turn a draft full of problems into two or three priorities a student can act on, written as principles that transfer to the next chapter.

Free. Works on Claude, ChatGPT, Gemini or any assistant that accepts a skill file.

What this skill does

The method, encoded.

Supervisory feedback fails in two conscientious ways. Returned with two hundred comments, it teaches nothing, because the student cannot see a hierarchy and so fixes the surface. Returned as "needs more critical analysis", it names a deficiency without naming its remedy. Underneath both sits the real tension: the feedback that most improves this draft is not the feedback that most improves this researcher.

This skill installs a method. Read the whole draft once without a pen and reconstruct its argument. Inventory every problem privately and sort it into blocking, structural, developmental and surface. Diagnose each one as a defect in the work or a gap in the student's understanding, which need different responses. Select two or three priorities on leverage, not severity, and write each as principle, located instance, consequence and action. Separate feedback on writing from feedback on thinking and tell the student which is which. Say early and in writing where progress is inadequate. Record what was advised, so transfer can be checked next round and a progression panel has something to read.

It is built for the academic's judgement, not in place of it: no mark, no progression call, and nothing sent unread.

Best used for

  • Returning a chapter draft with feedback a student can act on
  • Deciding which two of many visible problems to lead with
  • Converting a recurring correction into a transferable principle
  • Separating writing difficulty from thinking difficulty
  • Saying early and clearly that progress is not adequate
  • Building a supervision record for a progression review

Typical inputs

What you give it.

Student draft chapter, proposal or full thesis draft, The task the student was set and the stage of the programme, Marking or examination criteria the work will meet, Previous feedback given and the student's response to it, Time available before the next milestone, Institutional policy on AI use in supervision and on student data

Typical outputs

What you get back.

Prioritised feedback document with two or three priorities, Principle, instance, consequence and action for each priority, Explicit not-this-round list, Thinking versus writing diagnosis per issue, Reconstruction of the draft's argument as a check, Dated supervision record entry and next checkpoint

Method coverage

What the skill works through.

  1. Why feedback volume teaches the surface
  2. Reading the draft before annotating it
  3. Blocking, structural, developmental, surface: the four tiers
  4. Defect in the work or gap in understanding
  5. Choosing two priorities on leverage
  6. Principle, instance, consequence, action
  7. Writing problems and thinking problems are different faults
  8. Saying early that progress is inadequate
  9. The supervision record and what it is later needed for
  10. Where this skill stops and the institution takes over

Download

Free skill. One file.

Enter your email once. Every skill you download after that takes a single click.

How to install

Add the skill file and the five kernel protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them at the top of any assistant conversation. Then give it your real research material, not a description of it.

Download skill

Questions

Common questions.

How much feedback should I give on a student's chapter?

Two or three priorities, with everything else explicitly marked as not required this round. The constraint is not your time, it is the student's capacity to change several things at once. A draft returned with two hundred comments has no hierarchy in it, so the student fixes the surface and the structure survives untouched.

What should I write instead of "needs more critical analysis"?

Name the missing operation and where it is missing. For example: "you describe each study's findings but never say whether you find its method convincing; for the four studies you rely on most, add one sentence each on what its design can and cannot establish." The test is whether a competent student could act on the sentence without asking what you meant.

How do I tell whether a student has a writing problem or a thinking problem?

Try to restate the passage's claim in plain language. If you can, and the claim is sound, the problem is expression. If you cannot, or the restated claim collapses, the problem is the argument, and editing will not fix it. Tell the student which diagnosis you reached, because "this is unclear" leaves them unable to tell whether to rewrite the sentences or rethink the case.

When should I tell a student that their progress is not adequate?

As early as the evidence supports it, in writing, in its own dated message rather than buried in draft comments. A student told in month nine can recover; the same student told near submission cannot, and will reasonably ask why nobody said so. The message states the standard, what has been produced against it, what must change, and by when.

Can I use an AI system to help with feedback on a student's work?

Check your institution first. Using AI on student work may be restricted by policy, and it involves processing an identified person's personal data and unpublished work. Where the position is unknown, do not upload identifying or unpublished material. In every case the supervisor reads the draft themselves, and nothing reaches the student unread.

Can this skill mark the work or say whether the thesis will pass?

No. It never assigns a mark, never makes a progression or examination recommendation, and never issues feedback the academic has not read. Those decisions belong to the named academic and to institutional regulation. For assessment against stated criteria, the related skill is Thesis Examination and Marking.

My student keeps making the same mistake after I correct it. What am I doing wrong?

Probably correcting the instance rather than teaching the principle. A recurring error usually means the earlier feedback fixed a location rather than transferring a rule. Lead with the general principle, use one or two located examples as illustration, and check at the next submission whether it transferred to places you did not mark.

Research where people already are.
Analyse it where you already work.

Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.

New Report on SA Gambling Impact
Check It Out