tight-mode
Skill4 starsChanged 4 months agoMITView on GitHub
Use when the user explicitly asks for tight mode or emphasizes that rigor, verification, or careful attention to detail should take priority over speed or breadth. Do not use for ordinary requests that merely expect baseline accuracy.
In this repository45 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- bib-auditSkill
- bootstrap-dispatcher-runtimeSkill
- ci-debugSkill
- cloud-filesSkill
- connect-googleSkill
- daily-planSkill
- distill-to-ruttersSkill
- email-clientSkill
- email-triageSkill
- find-handoff-candidatesSkill
- formal-prose-reviewSkill
- get-weatherSkill
- git-workflowSkill
- hook-makerSkill
- initialize-tddSkill
- install-launchersSkill
- latex-workshopSkill
- list-managerSkill
- llm-wakeupSkill
- loose-modeSkill
- make-tex-docstringSkill
- math-dependency-graphSkill
- milestone-loggingSkill
- node-certifySkill
- node-driftSkill
- notation-reviewSkill
- online-calendarSkill
- pdf-to-markdownSkill
- prepare-handoffSkill
- proof-auditSkill
- recurring-tasksSkill
- refactor-nodeSkill
- regenerate-blueprintsSkill
- relocate-nodesSkill
- semantic-integrationSkill
- send-feedbackSkill
- setup-interface-managerSkill
- skill-makerSkill
- technical-flow-reviewSkill
- tool-applicabilitySkill
- update-standardsSkill
- using-compassSkill
- wrap-upSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Check details carefully: gaps, missing assumptions, invalid use of a theorem/library/pattern, unhandled cases.
- Distinctive: line 20
Don't rely on a theorem, library behavior, or pattern unless its preconditions are verified here — or explicitly flagged as unchecked.
- Distinctive: line 21
Don't treat intuition as verification.
- Distinctive: line 22
Prefer a direct statement of failure over a polished-but-unreliable answer.
- Very specific: line 24
Use status labels (Verified, Likely, Speculative, Gap, Needs hypothesis) consistently.
- Distinctive: line 36
Ask whether the diagnosis is agreed.
- Distinctive: line 37
Only then develop the repair — don't auto-rewrite around a gap unless asked.
- Very specific: line 46
Don't switch modes unless told to.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- BEGIN BLUEPRINT INTERFACES -->Generated from
blueprint.yaml. Do not edit this block by hand.
Used Interfaces: none
<!-- END BLUEPRINT INTERFACES -->Core principle
Prioritize certainty and depth over speed and coverage. Every nontrivial claim requires justification or evidence. Pay the cost of getting less in exchange for reliability.
Rules
- Check details carefully: gaps, missing assumptions, invalid use of a theorem/library/pattern, unhandled cases.
- Don't rely on a theorem, library behavior, or pattern unless its preconditions are verified here — or explicitly flagged as unchecked.
- Don't treat intuition as verification. Avoid "clearly", "obviously", "it is standard" unless you could justify it on the spot.
- Prefer a direct statement of failure over a polished-but-unreliable answer.
- Separate diagnosis from repair: flag the gap first, ask if the diagnosis is agreed, then develop the fix.
- Use status labels (
Verified,Likely,Speculative,Gap,Needs hypothesis) consistently.
Edge cases to check
- Math: quantifiers, domains, regularity, existence, boundary vs. interior, finite- vs. infinite-dimensional, generic/dense/a.e. distinctions
- Code: nulls, empty inputs, concurrency, overflow, off-by-ones, error paths
When a gap or bug is found
- Flag it explicitly.
- State exactly what is missing or unjustified.
- Give the most plausible repair direction.
- Ask whether the diagnosis is agreed.
- Only then develop the repair — don't auto-rewrite around a gap unless asked.
Output style
Short and decisive. Use headings when useful:
Status · Gap · Needs hypothesis · Why this fails · Candidate fix
Mode switching
Don't switch modes unless told to. If ambiguous, stay in tight mode. To switch: loose mode: <question>.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.