RussBrown00·RussBrown00/agent-skills/sniff-my-plan/SKILL.md

sniff-my-plan

Skill1 starsChanged 4 months agoGPL-2.0View on GitHub

Sniff-test a plan for INTENT ALIGNMENT via `/sniff-my-plan {file?}` — a light, pre-execution backstop that answers one question: do we actually know what we are building and why? Reads the plan (a file if given, otherwise the plan in context — one read/written this session, an inline plan in the conversation, or `.omc/plans/{TICKET}/{TICKET}.md` from the branch) and reconstructs its intent as a five-part statement — Goals, Why, How-it-achieves-the-goals, Non-Goals, and Success/Done criteria. Where the plan is too thin to fill a section, it ASKS — questions batched in small groups, grouped by section, each with a proposed default so the user can reply "defaults are fine." Once the statement is complete it starts an ALIGNMENT CONVERSATION: it states its understanding back, flags where the reconstructed intent and the plan's actual content diverge, and helps the user adjust. This is NOT roast-my-plan (that audits executability) and NOT dead-horse (that probes the agent's confidence): it is the lightest gate, checking comprehension and alignment BEFORE either of those. It is advisory and EPHEMERAL — screen-only, it never writes a file and never edits the plan. ACTIVATE when the user says "/sniff-my-plan", "sniff my plan", "sniff-test this plan", "what is this plan even for", "do we agree on the goal", "gut-check this plan", or "is this plan rich enough".