← the whole session plugin/skills/reflect/SKILL.md

Reflect on the ASSIGNED SCOPE that is still incomplete — what the human is trying to get, where the work stands against it, and the shortest path to close the gap. Never a reflection on the agent's own performance. Use after completing significant work, before declaring a task done, or at the start of a session to align on what you are about to do.

Reflection On The Assigned Scope — /reflect

Before your output, print ## AGENT_REFLECTION on its own line. This marks the block so it can be found again — by a session-end harvest step, by you on a later turn, or by any admin surface the person has built to read it. After printing, save it: append the target (quoted), the open gap, and the timestamp to a local file under alignment-harness records reflections (one file per session). This is what lets the same reflection be re-surfaced on a later message or after a compaction, instead of evaporating the moment the transcript scrolls past it.

THE SUBJECT OF THIS REFLECTION IS THE WORK, NOT THE AGENT

Read this before anything else. This is the single most common way this skill is run wrong.

The subject is the active scope the human assigned that is still incomplete. Every section below asks about THE WORK — what the human is trying to get, where the work stands against that, what closes the gap.

This is never a reflection on your own performance. Not your errors, not your process, not what you misunderstood, not how you drifted, not an accounting of your failures.

If you have made mistakes, they are relevant ONLY where they changed the state of the work — one clause naming what is now untrue or unbuilt, then back to the work. A section that narrates your psychology has failed. The human is not asking about you. They are asking about the thing they assigned and have not received.

Test before printing each section: is the subject of this sentence the work, or me? If it is you, rewrite it so it is the work.

If the assigned scope is large and you have completed only a fraction of it, the reflection is about the whole scope — including everything untouched. Reflecting only on the part you happened to work on repeats the same error at a smaller scale.

Think about this like a human would.

Before starting: Have I run the /governer skill? If not, run it first.

Steps

Work through each section. Use the number after the pipe as the target sentence count for your reply. Skip sections that lack clear signal — noise is worse than silence.

Compose the Reflection section last, then place it first in the response.

1. Target

What is the specific intent in UX terms the human is trying to get?

Do NOT leave out ANY explicitly stated user requirements. Also include translation of implied requirements into explicit UX statement requirements here at this step. What the user has stated they want, THAT is the target.

target | 1

2. Location

Where is the user in relation to that intent's potential?

location | 1

3. Gap

What is the gap they are trying to navigate?

gap | 1

4. Initiative

Why are they trying to cross that gap? What overarching intent can you deduce this aligns with? What would be the perfect representation of this potential fully realized? Think big at this point — this is not about the actual scope, it's about deriving the potential outcome that the scope could be seen as working towards, with explicit nuance and detail.

initiative | 1

5. Leverage

In navigating that gap, what have they not considered? What UX problems, gaps, considerations could help them? Harm them? What questions, if you asked yourself, would reveal the most elegant and graceful path to full realization of the initiative, and the most direct flow crossing the gap?

leverage | 1-3

6. Questions to Keep in Mind

What questions would you need to consider to support them in realizing how they could cross this gap? What questions matter for staying connected to the intent, the why, as you build? How could the why inform the build? What human-level checks would be required at completion to determine if you fulfilled the build?

questions to keep in mind | 1-3

answers to each question | 1

validation ux steps | 1-8

Score-Gated Additions

If this session has no real governer score yet (check alignment-harness status — it says "unscored" or names a provisional/baseline guess rather than a score you produced with /governer), treat that as the deep-question tier, not as license to skip them: a missing score is not evidence the work is low-stakes, and a baseline keyword guess must never quietly stand in for a real score to justify shallower reflection. Answer every score-gated block below by default.

If governer score is 60% or higher, reflect on each answer sequentially and print nuanced thinking.

These questions are about the STATE OF THE WORK. Answer them as facts about the scope, not as an assessment of yourself. "The intent has not been confirmed" is a fact about the work. "I failed to confirm the intent" is a fact about you — do not write the second form.

  • Is the work as it stands consistent with the intent behind it, or has it diverged?
  • Is the human's explicit intent written down anywhere in crystal clear UX terms?
  • Has the human confirmed that written intent is right, and when?

If 70% or higher, add:

  • Is the scope currently held in a decomposed state as discrete tasks, or only as prose?
  • Where does the work as it stands diverge from the intent, and in what direction?

If 80% or higher, add:

  • What edge cases could make this work break, fail silently, mislead the human into thinking it works when it does not, or break neighbouring systems?

If 90% or higher, add — the explicit gap between my intent to build from:

  1. Crystal clear intent translated to user UX target intent
  2. Self-eval of alignment with that intent
  3. Perfect intent alignment with clear intent
  4. Observability relative to the complexity of the task at hand
  5. Self-reflexivity and reasoning in relation to the task at hand
  6. Demonstration of capacity to discern the most effective and direct path towards the full realization of the client's intent
  7. Taking the graceful path through without losing context

Output

Focus on signal over noise. Skip sections that are not clear signal-rich. Aim for a nuanced, interconnected, and coherent reasoning summary at the end of the greatest signal you found.

Do NOT leave out ACTUAL VERBATIM REQUIREMENTS. Include them and STATE them explicitly.

For all score gradients, at completion compose the Reflection as a synthesis of all sections and place it FIRST in the response.

Reflection

A synthesis of the above — the through-line connecting target, gap, initiative, leverage, and questions into one coherent picture of what matters most right now and why.