downstream-instruction-framinglisted

# Downstream-Instruction Framing Discipline ## Problem When writing instructions for a downstream reader (handoff, plan, ADR, subagent prompt), LLMs default to authoritative framing: "Approved Decisions — Do Not Re-Litigate," "Execute the following steps," "These are final." This framing reads natural and decisive — and is almost always wrong. The downstream reader is engaged precisely because their proximity to the evidence (target repo, file access, ability to run tests, project-local rules loaded) gives them judgment the upstream author lacks. Command framing strips that judgment exactly when it is most valuable. The result: a smart downstream agent silently follows a plan that contradicts what they can plainly see in the code. The upstream author is **always** the less-informed reviewer relative to the downstream executor. Framing must reflect that asymmetry, not invert it. ## Context / Trigger Conditions Apply this skill when writing **any** of: handoff documents, execution plans, subagent dispatch prompts, workflow scripts, scheduled-agent briefs, ADRs proposing future work, or instructions to a fresh agent session opening elsewhere. Symptoms that it's being violated: headers like "Execute the following" / "Do Not Re-Litigate" / "These are final"; decision lists with no "Revisit if:" clauses; no acknowledgment that the downstream reader has more evidence access; no explicit license to disagree; imperative mood throughout ("Implement X. Do Y.") rather than proposa