Decide where to persist anything meant for future agent sessions, in any project. Use before saving a rule, preference, correction, workflow or fact about Pablo so agents won't need to ask or be told again ("remember this", "store this", "make sure this doesn't happen again"), and before adding or changing agent-facing guidance in any instruction file, skill, hook, context or reference doc, project map, brief or note. Not for ordinary code, human-facing docs, or session logs.
Installs into .claude/skills of the current project.
Are you the author of Instruction Placement?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/benthamite-instruction-placement)
---
name: instruction-placement
description: Decide where to persist anything meant for future agent sessions, in any project. Use before saving a rule, preference, correction, workflow or fact about Pablo so agents won't need to ask or be told again ("remember this", "store this", "make sure this doesn't happen again"), and before adding or changing agent-facing guidance in any instruction file, skill, hook, context or reference doc, project map, brief or note. Not for ordinary code, human-facing docs, or session logs.
user-invocable: false
---
# Instruction placement
1. Before choosing where anything goes, read the canonical policy in full:
`/Users/pablostafforini/My Drive/dotfiles/agents/instruction-placement.org`.
2. Apply its decision test and its permission boundary for `CLAUDE.md` and
`AGENTS.md`.
3. Name what will put the content in front of an agent at the moment it is
needed. If nothing loaded at that moment points to it, the placement is
incomplete.
Do not restate the policy here or elsewhere; it is the single source.
