Burn 451 CLI — manage your reading queue from the terminal. Brief, digest, triage.
Burn 451 CLI has 12 tools with basic naming and descriptions, but critical gaps in schema documentation and parameter descriptions. Tool names are action-oriented (login, save, delete, keep, read) but descriptions are brief (avg ~60 chars, well below the 194-char baseline). Most critically: input schemas are inferred from the rubric summary, not visible in source code provided. The 'triage' and 'digest' tools reference package.json as their source file, suggesting tool definitions may not be explicitly registered. No output schemas documented. Error handling is absent from visible code. Security concern: auth.mjs exposes SUPABASE_URL and ANON_KEY as constants, violating secret-injection pattern.
Daily launch page — what to read first, what's burning, what you absorbed today
Delete article from queue
Summary of what you've absorbed recently (moved to Vault)
Show articles burning soon (< 6h remaining)
Move article from Flame to Vault (permanent library)
List bookmarks by status (flame/vault/spark/all)
Set up authentication (get token from burn451.cloud/settings)
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., article IDs from list/search results needed for keep/read/delete).
Tools 'flame', 'stats', 'triage' have empty input schemas (no parameters documented). Cannot verify if they truly accept no input or if schemas are missing.
Descriptions are too brief (avg 57 chars vs 194-char baseline). 'Show articles burning soon' and 'Your reading triage statistics' lack context on WHEN to use, WHAT to expect, or HOW results are structured.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Mark article as read (moves to Spark)
Save URL to Flame (24h countdown starts)
Full-text search across all bookmarks
Your reading triage statistics
Interactive triage — k/r/d each article in Flame
Security: SUPABASE_URL and ANON_KEY hardcoded in auth.mjs (lines 8-9). These should be environment variables or server-side injected, not exposed in source. Violates secret-injection pattern.
No error handling guidance in tool descriptions. If 'save' fails (invalid URL, network error), LLM has no recovery path. Descriptions should include 'If URL is invalid, try...' or similar.