builders-diary 1.10.1
“Record what you actually judged in a work session — a local, judgment-first portfolio for AI-native builders. Installs a skill into Claude…”
Agent Skills SKILL.md definition fileAutonomous agent defaults past confirmations
SHA-256701627355bcd2b46b277a1222cf2f242ac0ba3d80b4cb6b492258c6a079d758d
MaleculeK(Am)
Evidence
30:48… orkflow**, even in the same conversation. Do not reuse
31a cached Skill body, helper path, project choice, candidate list, card approval, or prior dry-run
32output. Never say or assume “I already read the skill earlier.”
33
34Before any explanation, tool call, or decision:
35
361. Read the exact `SKILL.md` path supplied by th …
37 the path inside the Slash Command’s `<SKILL>` metadata. In Claude Code, this is the file that was
38 activated for the current invocation.
392. From that file, use only the absolute `python3 "..." --l …
40:27… e a helper from another AI client (for example, do not use
41 `~/.claude/skills/...` when the activated Skill lives under `~/.gemini/config/skills/...`).
423. If the exact activated file cannot be read, stop and report the missing path. Do not fall back to
43 a p …
96:45… The UI groups the derived toolbox into
97collapsible categories: Coding agent, Programming language, Framework / library,
98Data / backend, Deployment / infrastructure, and Research / validation.
99
100**highligh …
236:35… tion + card + evidence, one at a time
237
238For each confirmed task, run the following sequence before advancing to the next task.
239
240### A. Confirm this task's Purpose and Stage
241
242Use the client's structured question …
No evidence locations were recorded for this file. Raw result