Owner meta reasoning

Skill XHdW/owner-mode/skills/owner-meta-reasoning

Owner Mode for AI agents, implemented as Agent Skills: domain judgment, deliberate choice, and responsibility for work that must hold up in real use.

Install

npx -y skills add XHdW/owner-mode --skill owner-meta-reasoning

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

16 days oldThe repository was created 16 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

User-invoked higher-order reasoning that makes a provisional meta-level analysis itself the object, determines what that analysis should examine and how it should construct knowledge, then descends and applies the resulting analysis to the original task.

SKILL.md

3.8 KB, as published. Nobody here has run it

Owner Meta-Reasoning

Ascend so the lower level can see differently, then descend and use the difference.

Keep four relational objects distinct internally:

O: the object-level problem, phenomenon, artifact, or decision.
M1: a provisional meta-level analysis of how O should be understood, analyzed, evaluated, or made actionable.
M2: higher-order reasoning that makes M1 itself the object.
M1*: the retained, rebuilt, or composed analysis applied to O after the ascent.

The governing relation is:

O -> provisional M1 -> M2(M1) -> M1* -> O, evidence, and action.

This is meta-reasoning rather than statistical meta-analysis or a more elaborate pass through M1. Metrics and criteria may belong to M1; M2 determines the whole way M1 should construct knowledge about O.

Form the Meta-Level Object

Recover what understanding, judgment, or action the work on O must enable, who owns the values involved, and what reality can constrain the result.

Form the M1 that would ordinarily guide the work, but keep it provisional and unexecuted. Make legible the questions it would ask, the parts and level of O it would select, its categories and perspectives, its causal or logical model, admissible evidence, value choices, comparison frame, and boundary. M2 needs this actual analytical construction as its object.

Ascend and Determine M1*

Reason about M1 as a whole:

What must M1 make knowable, judgeable, or actionable to serve the purpose of O?
Which object selection, abstraction level, relations, categories, and perspectives make that possible?
What causal, logical, empirical, or normative structure can connect evidence to the needed conclusion?
Which evidence and warrants belong, and whose values determine relevance and tradeoffs?
Would a materially different M1 construct O differently enough to change understanding or action?

Determine what M1 should be, not merely whether the provisional M1 contains defects. Preserve what remains warranted; replace, combine, narrow, expand, or split what another construction serves better. M1* may retain the original form when the ascent establishes its fit.

Read references/deep-meta-reasoning.md when M1 combines several levels or epistemic aims, carries a consequential framing choice, or remains difficult to objectify as a whole.

Descend and Apply M1*

Execute M1* on O. Let proof, observation, experiment, use, or an actual artifact constrain the resulting understanding and action.

M1* must differ in consequence when it differs at all: it changes what is examined, how evidence bears on O, what conclusion can be supported, or what action follows. When M2 supports the provisional M1 unchanged, descend with it instead of manufacturing novelty.

Keep O and the user-authorized purpose visible while designing M1*. If the ascent reveals that O itself needs reframing, make that change explicit rather than silently treating it as a change to the analysis.

Ground the Ascent

One ascent from M1 to M2 usually supplies the needed higher-order control. Continue higher only when a newly surfaced question about M2 can materially change M1*; otherwise descend. The possibility of analyzing every analysis again is not itself a reason to do so.

Keep the level mechanics and private reasoning internal. Communicate the material change to the analysis and its object-level consequence when useful. A description of the hierarchy is not the result; the result is better understanding, judgment, or action on O.