SathiaAI

Adaptive terse-output mode that cuts filler, narration, preamble, and postamble from agent responses to reduce output-token burn — without ever cutting code, errors, decisions, explanations the human asked for, or verification steps. Use when the user says…

End-of-session ritual for AI working sessions — writes changelog with a session-economics footer (tokens, cost, contract score), updates status snapshot, scans staged diff for forbidden strings, checks audit trail integrity, writes a next-session contract…

Pre-flight procedure for writing and reviewing deployment runbooks that reference environment variables. Enforces git-grep-at-deployed-ref verification so runbook variable names match the code that will actually run. Covers finding the deployed SHA, grep…

Pre-publish scrub procedure for moving content from a private workspace to a public repository without leaking internal references, real pricing, employer names, or unverifiable claims. Covers file-level grep, claim softening, byte-verification after push,…

Safe discipline for landing multiple PRs through an automated merge gate. Prevents silent content loss from stacked merges, enforces hash-verification of target branch tips, and covers source-branch identification before pushing. Use when the user says "I…

Structured triage procedure for automated dependency-bump PRs (Dependabot, Renovate, or similar). Covers batching strategy, AI-reviewer false-positive handling, registry verification, and safe merge sequencing. Use when the user says "triage my dependency…

Manage an append-only decision log for AI-assisted projects, preventing silent reversals across sessions and models. Handles writing new entries, locking decisions, checking for conflicts with existing locked entries, and reading the ledger at session boot.…

Structured triage and rotation procedure for a leaked credential or API key. Enforces triage-before-revoke discipline to avoid turning a security incident into a production outage. Use when the user says "I leaked a key", "found a key in a commit", "API key…

Build and fix reliable multi-agent AI workflows for any skill level, from first-time builders to senior engineers. ALWAYS use this skill when the user asks to "build an agent", "automate" a task or workflow, wants "AI to do X for me", describes a multi-step…

Build and fix reliable multi-agent AI workflows for any skill level, from first-time builders to senior engineers. ALWAYS use this skill when the user asks to "build an agent", "automate" a task or workflow, wants "AI to do X for me", describes a multi-step…

Build and fix reliable multi-agent AI workflows for any skill level, from first-time builders to senior engineers. ALWAYS use this skill when the user asks to "build an agent", "automate" a task or workflow, wants "AI to do X for me", describes a multi-step…

Give an AI agent a verifiable identity and a tamper-evident record of what it did. Use this skill when the user asks to "badge an agent", "add VIA ID", "give my agent an identity", "make my agent's actions verifiable/auditable", wants to prove what an agent…

Independent adversarial review and deterministic release gating for code changes, using a multi-model reviewer panel over OpenRouter (or any OpenAI-compatible router, or an MCP transport like Composio). Use whenever the user asks for an adversarial review,…

Independent adversarial review and deterministic release gating for code changes, using a multi-model reviewer panel over OpenRouter (or any OpenAI-compatible router, or an MCP transport like Composio). Use whenever the user asks for an adversarial review,…

Check and update a project's tracked status (tracks, items, drift, decisions) in a self-hosted KnoTrack instance, without dispatching or orchestrating work.

Use this whenever the user is facing a choice or asks for a recommendation — "should I use X or Y", "which is better", "what should I do about...", picking a tool/vendor/library/approach, a purchase, an architecture call, or any "help me decide" moment, even…

Local-first KnoSky agentic GPS via MCP — Tier 0 map tools plus Mode B governed route/bundle/policy_check (DEC-108). Load when navigating a repo city, citing live files, or needing policy-checked routes. Not swarm-safe unless L3 doctor green.

SUITS — Understanding Intent Translator. Converts a raw human request into an AI-Ready Brief (true intent, goal, deliverable, audience, constraints, user-outcome success criteria, labeled assumptions) BEFORE work starts, asking at most ONE clarifying question…

Holistically review a pull request and remediate it. Synthesizes GitHub PR review comments, multi-model council or adversarial review outputs, issue tracker comments, and any pasted QA notes into a single remediation plan, finds adjoining issues the reviewers…

Use when designing a new LLM skill (Claude, GPT, Gemini, Llama, or any other), drafting a skill spec from an intent, or auditing an existing skill for hygiene. Triggers on requests like "design a skill that…", "spec out a skill for…", "write a skill that…",…

Frontier-grade reasoning discipline for any model or agent. Frames tasks with explicit success criteria, surfaces assumptions and tradeoffs instead of guessing, persists goals and decisions across long sessions in a SESSION.md state file, and verifies work…