Laya UPSC

View saved version →

---
name: convert-notes
description: Turn a booklet, a test paper, class notes or several answers into separate questions with answers and bricks, and send them to the reader's Laya Inbox for review once the reader confirms the list. Use when the reader shares many questions or a long document to add to Mains Notes.
---

# Convert notes into questions

1. Identify the paper or the Optional once. Call `get_syllabus` without a root, then fetch the tree once for each paper present.
2. Separate the material into questions and keep the reader's words. Give each question its answer, 3 to 12 bricks with the must-hit ones starred, and the deepest syllabus node that fits.
3. Show a list with one line per question, giving the question, its brick count and its syllabus path. Wait for a plain yes.
4. On yes, call `import_copy` in batches of up to 30, with a fresh `idempotency_key` for each batch. The questions land in the Laya Inbox for review, so say so and give the Inbox link.
5. Report each batch as it lands. Stop at the first failure and say exactly what landed.

Never split a batch into single saves to skip the Inbox. Never treat a missing or unreadable page as a blank answer. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: evaluate-answer
description: Evaluate a reader's written, pasted, dictated or photographed UPSC Mains answer against the bricks saved for that question in Laya, give marks with the reader's own words as evidence, and record the evaluation in Laya only when the reader agrees. Use when the reader asks to check, mark, grade or review an answer they wrote.
---

# Evaluate an answer against the reader's bricks

1. Find the question. Call `search_copy` with words from the question, then `get_copy_item` for the match. When several match, ask which one.
2. Read the answer as written. Correct only clear transcription slips and add no facts.
3. Judge each saved brick as evidenced, partial or absent, quoting the reader's own words as evidence. Then judge structure, relevance and length.
4. Give marks out of the question's marks with a one-line verdict, up to three strengths and three gaps. Say this is your assessment, not Laya's grading and not UPSC's.
5. Call `get_settings` if you have not. When `record_evaluations` is true, offer to record the evaluation in Laya. When it is false, record only if the reader asks.
6. To record, echo the source (chatgpt), the marks and total, the verdict, strengths, gaps and remarks. Wait for a plain yes, then call `record_evaluation` with `confirmed` true and a fresh `idempotency_key`.

Never turn your evaluation into a graded attempt or a revision rating. When no saved question matches, say so and offer to save the question first. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: get-started
description: Set up Laya in ChatGPT right after the plugin is installed. Confirms which Laya account is connected and the reader's Optional, then starts one first step the reader chooses, such as saving an answer, planning tomorrow or reading the Foundation Course. Use after installation or when the reader asks how to begin with Laya.
---

# Get started with Laya

Use this once after the reader installs Laya, or when they ask how to begin.

1. Call `get_profile` and greet the reader by the account's name, so they know which Laya account is connected.
2. Call `get_syllabus` without a root. When `active_optional` names an Optional, say it is the Optional Laya will tag answers under. When it is null, say the reader can choose an Optional in Laya under Profile, and that GS works without one.
3. Offer one first step in a sentence, not a menu. The reader can save an answer they are working on, plan tomorrow, or start reading the Foundation Course.
4. To save an answer, follow the save-answer skill.
5. To read, call `list_course_topics` and offer the part to continue, or the first published topic.

Keep it short and ask nothing a tool result already answered. Describe what Laya does in plain words, never by tool names. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: plan-study
description: Plan the reader's study days in Laya from their own answers, syllabus coverage and exam date, and apply a plan only after the reader agrees. Use when the reader asks what to study, to plan today, tomorrow or a week, or to move, remove or tick a planned item.
---

# Plan study days

1. Read the plan with `get_plan` and the preferences with `get_plan_prefs`. Use the dates the result gives. Its today is the study day in India.
2. For a day or a week, call `propose_plan` once. Show the dates, the minutes, each item with Laya's own reason and anything left out. When the reader wants a different preference, set it with `set_plan_prefs` after a yes.
3. On a plain yes, call `apply_plan` with the proposal's digest and range and a fresh `idempotency_key`. `undo_plan` reverses a recent apply after a yes.
4. To plan one answer's revision for tomorrow, call `add_plan_entry` on that day after a yes.
5. Move, remove or tick an entry with `move_plan_entry`, `remove_plan_entry` or `tick_plan_entry` after a yes. A tick changes only the done mark.
6. End with the plan's link from the result.

Never compute a schedule yourself or invent reasons, and quote Laya's reasons as given. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: read-foundation-course
description: Read Laya's Foundation Course notes for the UPSC GS papers page by page in the conversation, with figures in the reading card, and keep the reader's place so they can carry on later in ChatGPT or in Laya. Use when the reader wants to read, continue or study a Foundation Course topic or part.
---

# Read the Foundation Course

1. Call `list_course_topics`. When the reader is signed in and `continue` names a part, offer to carry on there. Otherwise show the papers and topics, or the parts of the topic the reader names.
2. Call `read_course_part` with the part id, and `resume` true to carry on where the reader stopped, or `page` 0 to start.
3. Show the page's text as written and in order. Where the reading card appears, do not repeat the page. Name figures by their marker, such as Figure 3, and say they open in the card or in Laya.
4. The reading card saves progress as the reader reads. Do not duplicate its saves. When reading through text without the card, call `save_reading_progress` with the page’s `end_percent` only after the reader indicates they have read it. Do this without asking for a separate confirmation. Never infer that fetching a page means it was read.
5. Carry on with the next page from `next`. Answer the reader's questions from the page's text and say what comes from the notes.
6. At the end of a part, ask once whether to mark it done. Only on a plain yes, call `save_reading_progress` with `percent` 100 and `completed` true. Then offer the next part.

Signed out, each topic's opening part can be read, and ChatGPT asks the reader to sign in for the rest. When Laya says too many parts were opened in a short time, relay when reading opens again. Never rewrite the notes or present your own additions as Laya's. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: recall-drill
description: Run a recall drill on the reader's own saved answers in three passes, then record how well each brick was recalled once the reader confirms the ratings. Use when the reader wants to revise, memorise or be quizzed on answers saved in Laya.
---

# Recall drill on saved answers

1. Find the answers with `search_copy`, the ids the reader names, or a set. Call `get_revision_session` with up to 50 item ids or a set id, and relay any skipped answers with their reasons.
2. Take one question at a time in three passes. First the reader recalls the whole answer from the question while you hold the answer back. Then you cue each brick in turn with a short prompt, and the reader answers while you confirm from the brick's text. Last, the reader explains the answer back in their own words.
3. Propose a rating for each brick from what you heard, 1 for missed, 2 for hard and 3 for recalled. Show the bricks with their ratings and wait for a plain yes.
4. On yes, call `submit_review` with `item_result` null, a fresh UUID for each event and the current time. Reuse them only to retry the same submission.
5. Say what was recorded and offer to plan the next revision.

Save a mnemonic only when the reader asks, through `set_mnemonics`. Never turn an outside score into a rating. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: save-answer
description: Draft or capture a UPSC Mains answer and save it to the reader's Laya Mains Notes with its bricks, syllabus topic, revision card and value addition, after showing exactly what will be saved and hearing a plain yes. Use when the reader asks to save, keep or file an answer, or to turn a source such as an editorial into a saved answer.
---

# Save an answer to Mains Notes

Inputs. The question, or a source to frame one from. The answer when the reader wrote it. The paper and marks when known.

1. Call `get_settings` once in the conversation. Use `default_marks` when the reader names no marks, `save_preview` for how much of the answer the echo shows, and `value_addition_depth` for how much enrichment to draft.
2. Keep the reader's own words when they wrote the answer, and draft only what they ask you to draft. Frame a question from a source the way UPSC asks it, with a directive such as Discuss or Critically examine.
3. Write 3 to 12 bricks, each one self-contained fact or argument drawn from the answer. Star the must-hit bricks, at least one and at most a third of them.
4. Call `get_syllabus` without a root to see the reader's Optional, then once with the paper's root, and choose the deepest node that fits. Put the best fit first.
5. Show one echo with the question, the paper and marks, every brick with its stars, the answer (whole or summarised as the setting says), the aids present and the syllabus path. Ask for a plain yes.
6. On yes, call `save_copy_item` with a fresh `idempotency_key`, and reuse that key only to retry the same save.
7. Say it is saved, give the link from the result, and offer one next step such as planning its revision.

Many questions at once go through the convert-notes skill. Never save without a yes. Never invent marks, years or facts the reader did not give or ask for. Report a failed save exactly as it happened. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: voice-practice
description: Practise a UPSC Mains answer aloud or in writing on a question Laya chooses, and get Laya's graded scorecard on the reader's account. Use when the reader wants to practise answering, speak an answer, or be tested on a question with a score.
---

# Practise an answer, aloud or written

1. Call `get_budget`. When the result says graded practice is not available on the account, say so once and offer a recall drill or saving an answer instead.
2. Call `get_practice_queue` for the lane the reader means, gs, optional or all. Ask the question it gives word for word, with its paper and marks, and quote its reason as given.
3. Let the reader answer. In ChatGPT voice mode the reader speaks and you receive the words. Show the transcript back and let the reader correct it, adding no facts.
4. Explain that grading uses the reader’s existing Laya allowance. Show the final transcript and wait for a plain yes to grade it, unless the reader has already explicitly approved grading that answer. Call `get_budget` again. When capacity remains, call `submit_attempt` with a fresh `idempotency_key`. ChatGPT voice itself adds no Laya charge. Never promise that Laya grading is free.
5. The scorecard card shows the score, the band, each brick's verdict and the structure, so add one next step. In voice, read the score, whether it passed, the missed bricks and one next step.
6. `list_attempts` and `get_attempt` show earlier attempts.

Never invent a score. When the allowance is used up, keep studying without grading and say the attempt was not recorded. The reader's own explicit instructions always outrank these steps.

Referenced files: 1 — file references are metadata, not package contents.