Content
77%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable archival skill with concrete SQL/CLI commands and well-sequenced validation checkpoints. Its main weaknesses are verbosity in the narrative Examples and failure to leverage the provided bundle files, leaving the SKILL.md monolithic.
Suggestions
Tighten the Examples into concise spec-style scenarios (table, criteria, batch size, outcome) rather than narrative paragraphs to improve token efficiency.
Move the archive-table DDL and batch SQL into assets/archival_template.sql and the automation logic into scripts/database_archival.py, then reference them from the body so the SKILL.md becomes a lean overview.
Remove generic prerequisite boilerplate ('Understanding of data retention requirements and compliance policies') that Claude already knows, keeping only operational requirements (credentials, CLIs, permissions).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The Instructions are dense with executable SQL and assume Claude's competence, but the three narrative Examples ('An orders table with 50M rows... VACUUM reclaims 72GB. Query performance on active orders improves by 40%.') and some generic prerequisite boilerplate ('Understanding of data retention requirements and compliance policies (GDPR, HIPAA, SOX)') could be tightened without losing clarity. | 2 / 3 |
Actionability | Fully executable commands throughout: `CREATE TABLE orders_archive (LIKE orders INCLUDING ALL)`, `INSERT INTO orders_archive SELECT *, NOW(), batch_id FROM orders WHERE <criteria>`, `COPY (...) TO '...' WITH CSV HEADER`, `gzip`, and `aws s3 cp ... --sse aws:kms` are copy-paste ready with only justified criteria placeholders. | 3 / 3 |
Workflow Clarity | A clear 10-step sequence with an explicit validation checkpoint in step 5 ('Verify row counts match' then 'Delete from source only after verification') wrapped in a transaction, plus an Error Handling table providing feedback loops such as 'restart from the last completed batch using batch_id tracking' and 'retry with exponential backoff'. | 3 / 3 |
Progressive Disclosure | The body is well-sectioned internally (Overview, Prerequisites, Instructions, Error Handling, Examples) but is effectively monolithic: the provided bundle directories (assets/, scripts/) contain only placeholder READMEs and are never referenced from the body, and content that could be split out (SQL templates, automation scripts) is inline rather than in the bundled files. | 2 / 3 |
Total | 10 / 12 Passed |