| name | study-it-out |
| description | Ground analytical judgment in the actual artifact before rendering it. Load before producing OR dispatching a review, critique, evaluation, assessment, or "is this right / this is the bug" verdict — anywhere a conclusion could be pattern-matched from a role instead of read from the thing. Named for D&C 9:8. |
Study It Out
"you must study it out in your mind; then you must ask me if it be right." — D&C 9:8
The failure this prevents
A reviewer with no grounding produces output that is structurally correct but factually invented — a review that reads right without having read the artifact. A cold-start subagent is most prone to it: given only a role and a folder path, it pattern-matches from its role definition and fabricates a plausible verdict. But the orchestrating session does it too, under time pressure, when it "reviews" from memory of what the code probably says.
This is read_before_quoting extended from quotes to judgments. Close-enough wording is fabrication; a close-enough review is fabrication too.
The discipline
Before you render — or dispatch — any analytical verdict (review, critique, eval, assessment, "looks good", "that's the bug"):
- Study it out first. Read the actual artifact this session — the real diff, the real file, the real log. Not your memory of it, not its name, not its role description.
- Cite or it didn't happen. Every finding references a specific
file:line / section. An ungrounded review can't cite real lines — so an uncited verdict is the fabrication smell. N findings → N references (the cite-count rule, applied to reviews).
- Confirm the artifact is real before reviewing it. Missing or placeholder input is the finding — don't review around a void.
- Keep apex discernment with the context-holder. The whole-picture judgment ("does this serve the intent?") stays with whoever holds the whole arc. Dispatch grounded narrow checks, not the judgment that needs full context.
- If you must hand off, equip the witness. Package the real artifacts (full paths, the procedure, the work item) so the next session studies it out instead of starting cold.
When it applies / when it doesn't
- Applies to analytical output: reviews, critiques, evaluations, source assessments, "is this right?" verdicts — yours or a dispatched agent's.
- Does not gate grounded, verifiable execution: file moves, code edits that will be diffed, search results, build/test runs. Those are safe to dispatch cold — their output is checkable by inspection. The line is judgment vs. execution — the same line as "delegate execution, not discernment."
The tell
If you can't point at the line, you didn't study it out — you guessed. The burn is counterfeit when the study was skipped.
It cuts both ways
Unverified skepticism is as ungrounded as unverified credulity. Flagging a claim as "tradition," "unsupported," or "presented as fact" without checking the source is the same failure as asserting it without reading — you've rendered a verdict from your prior, not the artifact. Verify before you rule, in either direction. (Learned live, 2026-06-07: a yt-gospel eval flagged the Wilford Woodruff St. George founders account as "⚠ tradition"; the user pushed; the Church's own 1991 Ensign Q&A and Woodruff's journal documented it plainly. The skeptical ruling was the ungrounded one.)
Relation
- Extends the read-before-quoting discipline from quotes to judgments.
- The same discipline can be built into automated pipelines as a hard gate: no verdict ships until the evidence of actually reading the artifact is attached.
- Sibling of the full-context-shepherd lesson and the critic architecture.
