writing-for-agentslisted
Install: claude install-skill praisethebit/skills
Reference for any document an agent reads: a skill, an `AGENTS.md` / `CLAUDE.md`, a doc reached by a pointer. The same levers make each one predictable. The agent follows the same _process_ each run.
For a skill, also read [`SKILL-MECHANICS.md`](SKILL-MECHANICS.md): frontmatter, invocation choice, router skills.
## Context pointers
A **context pointer** is a reference in the agent's context that names out-of-context material and the condition to reach it. A skill description is one. A line in `AGENTS.md` that names a doc is one. The pointer's _wording_, not its target, decides when and how reliably the agent reaches the material. If a must-have target sits behind a weak pointer, sharpen the wording first. Inline the material only if that fails.
A pointer states what the material is and lists the **branches** that trigger it (a branch is a distinct case, so runs take different paths). An always-loaded pointer costs tokens on every turn, so prune it hard:
- **Front-load the leading word.**
- **One trigger per branch.** Merge synonyms of one branch.
- **Cut identity the body already carries.**
## The two loads
- **Context load**: always-loaded material (an `AGENTS.md` line, a skill description) costs tokens and attention every turn, whether or not it fires.
- **Cognitive load**: the human must know which documents exist and when to use each. Spend it where human judgement matters. Remove it where it does not.
Material behind a pointer costs only the pointer line. Materia