diff --git a/AGENTS.md b/AGENTS.md
index e329431..04ce200 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -38,3 +38,11 @@ Scope: Applies to the entire repository unless overridden by a deeper AGENTS.md.
 - Ask clarifying questions in `collab/<...>-questions.md`.
 - Propose in `collab/<...>-proposal.md`, plan in `collab/<...>-plan.md`, then implement.
 - Respect repository structure and this AGENTS.md across all changes.
+
+## Token/Context Hygiene
+- Chat minimalism: keep messages short; prefer pointers to files over inline dumps.
+- Use `.llm.md` files for concise, machine-friendly summaries; avoid duplicating full human docs.
+- No large pastes in chat: never paste large code or markdown; write to files and reference paths/sections.
+- Incremental artifacts: start a new `NN-<subject>.md` per round to avoid long single files.
+- Status style: one-line updates before tool calls; rely on devlog for detail.
+- Avoid re-reading large files in chat; reference exact file paths and sections instead.
diff --git a/SeedPrompt.md b/SeedPrompt.md
index f017c7c..1f05d07 100644
--- a/SeedPrompt.md
+++ b/SeedPrompt.md
@@ -61,3 +61,10 @@ Commit/push everything in the repo now before we get started on the project.
 - Use multiple files per round rather than growing a single document.
 - Naming: `collab/<type>/NN-<subject>.md` (human) and `collab/<type>/NN-<subject>.llm.md` (LLM-optimized).
 - Start a new `NN` for each major round; keep subject consistent across types.
+
+### Token Discipline
+
+- Keep chat output extremely concise; use the filesystem for details.
+- Summarize for the model in `.llm.md` companions; keep human-readable detail in the primary `.md`.
+- Do not paste large content into chat; write to files and reference paths.
+- Prefer starting new numbered files per round to keep each doc short.