hi-claude
01Plugin Claude Code
Hi Agencia's plugin marketplace — home of hi-claude, the best-practices working method for Claude Code.
Plugin Claude Code
Hi Agencia's plugin marketplace — home of hi-claude, the best-practices working method for Claude Code.
Plugin Claude Code
A working method for Claude Code: seven principles on four axes with the hierarchy that settles them, governed memory, a curated CLAUDE.md, a register of open work that survives compaction, every content with a declared destination instead of a new file, text that gets pruned when it stops being true, subagents that…
Instructions file
Claude Code instructions for HiAgencia/hi-claude, covering hi-claude (plugin source), start here, how to work here — the hi-claude method, project structure and gotchas — what the tree does not say.
Agent
Audits a project's CLAUDE.md against the hi-claude method rubric (no inline docs, every line still true and still useful to a future session, tools table, structure section, timeless rules, no dead paths, no secrets). Line count is reported and scores nothing. Use when the user asks to audit or improve CLAUDE.md, or…
Agent
Audits the project's documentation for CURRENCY — text that stopped being true or stopped serving a future session: executed plans still presented as open work, documents the index declares current while a newer one exists, versions below the shipped one, chronicles of what happened. Use when the user asks whether the…
Agent
Audits the project's inventory document (docs/INVENTARIO.md / docs/INVENTORY.md, or the older split SKILLS/MCP/PLUGINS/TOOLS files) against what is actually installed and connected — drift both ways, proportional detail, capabilities claimed without evidence, and whether one subject is still split across four files.…
Agent
Audits the project's persistent memory directory (/.claude/projects/ /memory/) against the hi-claude protocol — admission rule, canonical format, index health, duplicates, oversized memories, secrets. Use when the user asks to audit, clean, or review memory, or via the hi-claude audit skill. Examples: Context: User…
Agent
Audits the project's folder organization against the hi-claude method — clean root, docs/ with INDEX.md, brand assets in BRAND/, no stale/temp files, no plain-text secrets, real structure matching what CLAUDE.md declares. Also audits BELONGING: documents that exist and no CLAUDE.md declares, pairs that should be…
Agent
Audits docs/ROADMAP.md and docs/ESTADO.md against the hi-claude ROADMAP protocol (no closed items left behind, no dates, open-work block present and capped, progress lines that did not become a log, runnable done criteria). Use when the user asks to audit or clean the roadmap or the pendings, or via the hi-claude…
Skill Claude CodeCodex
Run hi-claude audits over the project. Use when the user asks to audit, review, check, or clean the project, its CLAUDE.md, its memory, its roadmap, its organization, or whether its documentation is still true ("auditá", "revisá el proyecto", "limpiá esta carpeta", "¿esto sigue sirviendo?", "está lleno de cosas…
Skill Claude CodeCodex
Use BEFORE dispatching any subagent, agent or parallel investigation, and again the moment a subagent's report comes back — under this method a subagent never decides what gets done, and what it returns is a hypothesis nothing may rest on until it is verified first-hand. Triggers on "delegá esto", "mandá un…
Skill Claude CodeCodex
Governs what enters persistent memory and CLAUDE.md under the hi-claude method. Use whenever the user corrects an approach, confirms a non-obvious decision or a validated approach ("yes, exactly", "sí, exactamente", "perfecto, así"), states a preference or a limit, or says things like "remember", "always", "never do"…
Skill Claude CodeCodex
Use at the start of a session to recover what was in progress, when taking, splitting, pausing or closing a piece of work, when the user wants something recorded so it does not get lost, and before a context compaction — even when they only ask you to note it down. Triggers on "what were we doing", "where did we leave…
Skill Claude CodeCodex
Use whenever quality stalls, something feels off without a clear cause, the user says it cannot be done, or a big block of work just closed (inverse pre-mortem) — even when they only ask you to look again. Triggers on "algo anda mal y no sé qué", "no se puede", "estamos estancados", "qué se nos está escapando"…
Skill Claude CodeCodex
Day-1 onboarding for any project under the hi-claude method. Use when the user wants to set up, initialize, or organize a project ("set up this project", "configurá el proyecto", "empecemos", "cómo arranco", "ordená esta carpeta"), when a folder has no CLAUDE.md and the user asks how to begin, or right after the…
Skill Claude CodeCodex
Use right after the hi-claude plugin is updated, and whenever the user asks what a new version brings or whether their project has to change because of it. Triggers on "actualicé el plugin", "actualicé hi-claude", "qué trae la nueva versión", "¿tengo que cambiar algo?", "¿me falta aplicar algo?", "¿esto quedó viejo?"…
Skill Claude CodeCodex
Use BEFORE writing, updating, MOVING, MERGING or DELETING any documentation, README, code comment or commit message — this project BANS specific patterns in written work (dated state, verdicts written as final, docs that turn into a chronicle), decides WHERE each content belongs before it is written, and carries the…