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.
A well-built diagnostic skill body: executable commands, precise tool parameters, explicit pass/fail criteria, and a clear run-all-then-summarize workflow with error-recovery guidance. Its main defects are the undefined '17 known coding categories' list (which blocks one check from being executed) and mild verbosity/duplication around the identity and session checks.
Suggestions
List the 17 known coding categories (inline or in a references file) so Quality Check 4's orphan-memory rule is actually executable — currently the check references a list that appears nowhere in the skill.
Consider moving the Extended mode (--deep) memory quality analysis into a separate reference file (e.g., references/quality-checks.md), keeping SKILL.md focused on the 5 core checks.
Trim the Check 2 rationale paragraph ('the plugin already resolved branch and project from git at session start...') to a one-line instruction ('Report MEM0_BRANCH verbatim; do not re-run git'), and deduplicate the env-var echoing shared between Checks 2 and 5.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean — commands, pass/fail criteria, and output templates — but the Check 2 paragraph explaining why not to re-run git over-explains, and Checks 2 and 5 duplicate echoing MEM0_APP_ID/MEM0_BRANCH, so a few tokens could be trimmed without losing clarity. | 4 / 5 |
Actionability | Mostly fully executable: copy-paste bash snippets, exact tool-call parameters (filters, top_k, infer=False), and per-check PASS/WARN/FAIL criteria. However, Quality Check 4 instructs flagging memories 'not in the 17 known coding categories' without ever listing those categories, leaving that step non-executable and keeping it below the copy-paste-ready anchor 5. | 4 / 5 |
Workflow Clarity | Explicit sequencing ("Run ALL checks, then display a single summary. Do not stop on the first failure"), explicit validation checkpoints per check, async event-status polling with a PENDING-after-5s rule, probe cleanup via delete_memory, and a conditional Troubleshooting section that creates a fail-to-fix feedback loop — matching the anchor-5 example structure. | 5 / 5 |
Progressive Disclosure | No bundle files exist and the main body is well organized with clearly gated extended mode ("When invoked with --deep"), but the ~60-line quality-analysis block is a candidate for a separate reference file and the dangling '17 known coding categories' allusion points to detail that is documented nowhere — minor organization gaps matching anchor 4. | 4 / 5 |
Total | 17 / 20 Passed |