Repository navigation
Skill Catalog
SternOne edited this page Sep 23, 2026
·
12 revisions
There are seventeen skills: the epistemic usage guide and sixteen methods. The groups below match the method map on How the Pieces Fit and only make the list easier to scan. You can use any method on its own, in any order.
| Method | The question it helps answer | What you get back |
|---|---|---|
| Epistemic | How do I use this suite on this task? | The agent loads the instructions that apply and tells you briefly what each method added |
| Method | The question it helps answer | What you get back |
|---|---|---|
| Metacognate | Is our reasoning aimed at the right outcome? | A corrected assumption, a better check, a clearer statement of confidence, or a note that the approach holds |
| Recon | What are we missing about this task before we start? | A corrected brief, a map of the decisions ahead, or a record of what an outside project could offer |
| Resolve | What kind of evidence could settle this question? | An answer or a named limit, worked out by reasoning, found in published research, or tested with a small throwaway probe |
| Open Questions | Which decisions belong to the user? | An interview, then a list of what was decided, deferred, dropped, or is still on hold |
| Method | The question it helps answer | What you get back |
|---|---|---|
| Perspective | What does one focused angle show, such as whether this can be undone? | A supported finding or improvement, a confirmation that the approach is adequate, or a named gap in the evidence |
| Gauntlet | Does an important decision hold up when several separate reviews examine it? | GO, CONDITIONAL, or NO-GO, with disagreements kept and the limits of the review stated |
| Method | The question it helps answer | What you get back |
|---|---|---|
| Health | What state is the system in right now? | OK, WARN, CRITICAL, or UNKNOWN for each thing checked, with what was seen |
| Triage | What caused this failure? | A verdict (cause found, narrowed, unknown, or not broken) with the observation that tells the explanations apart |
| Did It Land | Did the change take effect where it is used? | LANDED, REVERTED, or UNVERIFIED, for the place and time that were checked |
| Watch | Who will notice if a limit is crossed after this session ends? | A record of where the watch stands. It is marked PROVEN only after a running check has sent a safe test alert through the real path to the person who agreed to receive it |
| Evidence-Locked UAT | Does what the user sees and does meet criteria set before testing? (UAT is user acceptance testing.) | An acceptance result tied to what was observed, with what was left out |
| Method | The question it helps answer | What you get back |
|---|---|---|
| Write Goal | What counts as done, and can your agent's app track it as a goal? | A written definition of done; a goal is switched on only when you allow it |
| Decision Ledger | Can the next session find this reasoning and recheck it? | One record the next session can use, reusing an existing one when it is enough |
| Manifest | Does this long job need a formal record that survives interruptions? | A mission record the next session can resume from; only someone other than the agent that did the work can accept it as done |
| Outsource | Can another agent or tool pick this up without this chat? | A fixed handoff packet it can work from, a check of whatever comes back, or — when ownership itself moves — a recorded acceptance and a completed handover |
| Context Audit | Are the agent's loaded instructions conflicting or stale? | A list of conflicts or duplicates and a proposed fix you can review |
For close calls between similar methods, see How the Pieces Fit. For worked examples, see Workflow Recipes.
Applies to: epistemic-skills v7.3.0. Release notes · Source for these pages · The skills as released
Start here
Use the skills
Understand the design
Install and troubleshoot
Maintain and verify