Content
81%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body delivers fully executable commands, a clearly sequenced escalation workflow with explicit stop conditions and an error-recovery fallback, and useful output-field documentation. Its main weakness is redundancy — the skim-is-abbreviated concept and the no-plain-git rule are each repeated several times across sections, inflating token cost without adding guidance. The field listings could live in a reference file to shorten the main body.
Suggestions
State the 'skim is complete-but-abbreviated, not a full transcript' concept once (e.g., in Reading `skim`) and delete the restatements in Core Workflow, the numbered steps, Rules, and Skim Summary.
Consolidate the 'do not run plain Git commands first / do not use gitbutler/workspace as a target' guidance into a single Rules bullet instead of three separate passages.
Move the `skim` and `show` field lists into a references file (e.g., references/fields.md) and reference it one level deep, trimming the main body toward a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean and imperative, but the core idea that 'skim is complete-but-abbreviated' is restated roughly five times (Core Workflow, the numbered steps, Rules, Reading `skim`, and Skim Summary), and the prohibition on plain Git commands appears three times. These repetitions could be cut without losing clarity. | 3 / 5 |
Actionability | Commands are copy-paste ready and cover the common cases: `but agentlog skim`, `but agentlog skim branch <branch-name-or-ref>`, `but agentlog show <session-key> --limit 20`, `but --json agentlog skim` for handles, and `but status` as the discovery fallback. Field lists for skim/show output make drill-down concrete. | 5 / 5 |
Workflow Clarity | The workflow is clearly sequenced with explicit checkpoints and feedback loops: numbered escalation (skim → JSON handles → show session → show turn), stop conditions ('If `skim` is enough... summarize it and stop'), and an error-recovery path ('If target discovery fails, then inspect GitButler state: `but status`'), plus bounded `--limit` escalation rules. | 5 / 5 |
Progressive Disclosure | Well-organized single-file skill with clear sections (Core Workflow, Target Discovery, Rules, Reading `skim`, Reading `show`, Skim Summary) and no broken references, but at ~175 lines the inline field-reference lists push it past the lean single-file ideal; no bundle files exist to evaluate. | 4 / 5 |
Total | 17 / 20 Passed |