/ase-meta-review

name: ase-meta-review
argument-hint: "[--help|-h] [--severity|-S=(LOW|MEDIUM|HIGH)]"
description: >
    Perform a holistic, human-reviewer-style critique of the currently
    staged Git changes and emit an approve/reject verdict with
    prioritized, severity-tagged, line-cited findings. Use when the user
    wants the staged diff "reviewed", "critiqued", or "code-reviewed"
    before committing.
user-invocable: true
disable-model-invocation: false
effort: high
allowed-tools:
    - "Bash(git diff *)"
    - "Agent"

@${CLAUDE_SKILL_DIR}/../../meta/ase-control.md @${CLAUDE_SKILL_DIR}/../../meta/ase-skill.md @${CLAUDE_SKILL_DIR}/../../meta/ase-getopt.md

<purpose name="ase-meta-review"> Review Staged Changes </purpose>

<expand name="getopt" arg1="ase-meta-review" arg2="--severity|-S=(LOW|MEDIUM|HIGH)"> $ARGUMENTS </expand>

<objective> Review the currently staged Git changes the way an *experienced human reviewer* would - judging them *holistically* against the change's *own intent* and against *correctness*, *design fit*, *clarity*, *robustness*, and *project-convention conformance* - and emit a single *approve / request-changes verdict* backed by *prioritized*, *severity-tagged*, *line-cited* findings. This is a *synthesizing critique*, not a mechanical scan: it complements `ase-code-lint` (mechanical quality), `ase-code-analyze` (logic/semantics), and `ase-meta-diff` (intent narrative and risk). </objective>

Procedure ---------

<flow>

1. <step id="STEP 1: Determine Change Set">

1. Determine *whether there are staged changes at all* by running the corresponding command (taken exactly as given) and capturing its output - the bare *list of staged file names* - into <diff/>. This is a lightweight gate; the full diff is fetched by the sub-agent in STEP 2, so capturing only the file-name list here is sufficient:

`git diff --cached --name-only HEAD`

2. <if condition="<diff/> is empty"> Only output the following <template/> and then *STOP* immediately:

<template> ⧉ **ASE**: ✪ skill: **ase-meta-review**, ▶ status: **no changes to review** </template> </if>

</step>

2. <step id="STEP 2: Review Investigation">

<if condition="<ase-project-boxing/> is equal `black`">

The project source artifacts are classified as a *black box*, so the user does *not* want the staged changes scrutinized or their findings surfaced. *Skip* the entire review investigation: do *not* invoke the `Agent` tool and do *not* read any change, set <findings/> to the *empty* list, set <verdict/> to `SKIPPED (boxing: black)`, set <summary/> to a *one-line* neutral restatement of the change intent derived solely from STEP 1, and proceed *directly* to STEP 3.

</if>

First, use the following <template/> to give a hint on this step:

<template> <ase-tpl-bullet-secondary/> **REVIEW INVESTIGATION** </template>

Dispatch the review investigation to a *sub-agent* via the `Agent` tool so that *no* investigation details leak into the user-visible transcript. The sub-agent performs the silent reading, the read-only repository probing, and the critique; only its final structured return value is consumed here.

For this, invoke *exactly once* the tool:

        Agent(
            description:       "Review Investigation",
            subagent_type:     "ase:ase-meta-review",
            prompt:            "Review the staged changes.",
            run_in_background: false
        )

Parse the single result message of the `Agent` tool as a JSON object, set <summary/> to its `summary` field (a single crisp sentence reconstructing the change's intent), and set <findings/> to its `findings` field (a list).

Then *derive* the overall <verdict/> from <findings/>: set <verdict/> to `REJECT - DEMANDS CHANGES` if *any* finding in <findings/> has a `severity` field of `HIGH`; otherwise set <verdict/> to `APPROVE`. The verdict is derived *before* the severity floor below, so the floor only affects which findings are *rendered*, never the verdict.

Then determine the *effective severity floor* <floor/>: define the ordinal rank `LOW`=1, `MEDIUM`=2, `HIGH`=3, start from <floor><getopt-option-severity/></floor> (default `LOW`), and - if <ase-project-boxing/> is equal `grey` - raise <floor/> to `MEDIUM` whenever its current rank is below `rank(MEDIUM)` (grey boxing surfaces only *material* findings of severity `MEDIUM` and above). The floor affects only which findings are *rendered*, never the <verdict/> derived above.

Then *apply the effective severity floor* <floor/>: *Keep* a finding in <findings/> if and only if its `severity` field is `ACCEPTED` *or* `rank(severity)` is greater than or equal to `rank(<floor/>)`; *silently drop* all other findings. With the default floor `LOW`, all findings are kept. `ACCEPTED` findings are *never* dropped.

You *MUST* *NOT* output anything else in this STEP 2.

</step>

3. <step id="STEP 3: Verdict and Findings">

1. Use the following <template/> to output the overall review in <verdict/> and the reconstructed intent <summary/>:

<template>

<ase-tpl-bullet-signal/> **REVIEW VERDICT**: **<verdict/>**

<ase-tpl-bullet-normal/> **CHANGE INTENT**: <summary/>

</template>

You *MUST* *NOT* output anything else in this STEP 3.1.

2. <if condition="<findings/> is empty"> Only output the following <template/> and then *SKIP* the remainder of this STEP 3 (an empty finding list implies the `APPROVE` verdict, so the closing hint of substep 4 would expand into nothing anyway):

<template>

<ase-tpl-bullet-normal/> **NO FINDINGS**: the change is clean, nothing to flag.

</template> </if>