Prompt components
Prompt components are small add-ons you paste into a user message template (.user.txt).
They are not standalone workflows.
Quick picks
- Prevent automatic agreement with strong user assertions → anti-auto-agreement (user)
- Ensure complete reading of user-provided artifacts → deep-read (user)
- Require exhaustive artifact scanning + coverage reporting → deep-scan (user)
- Require web retrieval + citations (when tools are available) → deep-search (user)
- Add a structured analysis/compliance pass → deep-analyzed (user)
Component catalog
| Component | Adds | Use when | File |
|---|---|---|---|
| anti-auto-agreement (user) | Prevents automatic confirmation of user assertions without evidence | You want stance-neutral answers (no “you’re right” mirroring) | anti-auto-agreement.user.txt |
| deep-read (user) | Read all provided artifacts before answering | Output depends on user-provided text/files | deep-read.user.txt |
| deep-scan (user) | Exhaustive artifact scan + coverage reporting + evidence pointers | You have repo files / uploaded attachments and need completeness before conclusions | deep-scan.user.txt |
| deep-search (user) | Web retrieval + citations for external/up-to-date facts | You need current/public facts and browsing/tools are available | deep-search.user.txt |
| deep-analyzed (user) | Structured analysis + missing-input/contradiction/compliance checks | Reviews/debugging/comparisons where depth matters | deep-analyzed.user.txt |
Usage rules
- Paste components into a user runner (
.user.txt) unless explicitly documented otherwise. - Use 1–2 components max per run to reduce instruction collisions.
- Components that require tools (e.g., browsing) should be used only in runtimes that support those tools.
anti-auto-agreement (user)
Adds: prevents automatic agreement/confirmation of user assertions unless supported by evidence.
Use when: you need objective answers and want to avoid stance mirroring.
Avoid when: there is no user stance/assertion to evaluate.
File: anti-auto-agreement.user.txt
deep-read (user)
Adds: complete reading of user-provided artifacts before answering.
Use when: the task depends on provided text/files.
Avoid when: there are no artifacts/inputs.
File: deep-read.user.txt
deep-scan (user)
Adds: exhaustive artifact scanning + explicit coverage disclosure + evidence pointers for non-trivial claims.
Use when: you have multiple artifacts (repo/zip/files) and need high confidence you didn’t miss relevant content.
Avoid when: there are no artifacts to scan.
File: deep-scan.user.txt
deep-search (user)
Adds: explicit web retrieval + citations for external/up-to-date facts.
Use when: correctness depends on current/public sources and browsing/tools are available.
Avoid when: artifacts-only mode OR tools are unavailable.
File: deep-search.user.txt
deep-analyzed (user)
Adds: a structured analysis pass + explicit checks (missing inputs, contradictions, output compliance).
Use when: reviews/debugging/comparisons where depth matters.
Avoid when: trivial tasks where the overhead adds latency without value.
File: deep-analyzed.user.txt