3.2.0-beta.10). Features may change before the next stable release.Switch to stable →ak skills self-improve review
ak skills self-improve review inspect
Show one screened skill with its qualifying reasons, the evidence event ids for ak-dogfood, its decision, its candidates, and the next commands.
Use ak skills self-improve review inspect to look at one screened skill in
detail before you decide. The command is read-only.
Usage
ak skills self-improve review inspect <skill> [flags]ak skills self-improve review inspect ak-fix
ak skills self-improve review inspect ak:fix --jsonExactly one skill reference is required. It is matched case-insensitively,
with : read as - and any publisher or kit prefix dropped. The command has
no flags of its own. Shared flags are described in
CLI conventions.
What it does
The command shows the same item as review list, plus each
qualifying reason and the evidence event ids to pass to
/ak-dogfood --evidence. Use
ak skills evolution show <id> to see a
candidate's diff.
It reads the same files as review list, writes nothing, and runs no model,
script, or network call.
Output
The text output adds one reason line per qualifying problem, with its
fingerprint, signal type, error class, strength, and session count, and an
evidence line with the event ids.
JSON uses kind=skills.self_improve.review.inspect. data is one item with
the fields that review list documents.
Exit behavior
| Exit | Meaning |
|---|---|
0 | Success. |
1 | The AgentKit home could not be resolved (self_improve.home_unresolved), or the event store could not be read (self_improve.store_failed). |
2 | Invalid flags or arguments (self_improve.invalid_input). |
4 | Another writer held the event store lock (self_improve.store_busy). |
5 | The skill has no screened events (self_improve.not_found). |
Related commands
ak skills self-improve review defer
Defer the review of a screened skill, silencing its session-start line for 7 days by default and at most 30.
ak skills self-improve review list
List every screened skill with its latest decision, its recorded improvement candidates and their evaluation outcome, and the next commands.