omp-pantheon

Initialize a foreign project to use the pi-seshat system. Symlinks .omp/, drops AGENTS.md and CLAUDE.md templates, scaffolds specs/ and .pi/, updates .gitignore. Dry-run by default; --i-approve gates all mutations.

Read-only cross-source consistency checker. Surfaces drift between Linear issue state, SpecSafe slice files, and commit Spec-Slice trailers. No mutations, no network writes.

Propose-review-apply flow for BMad artifacts (PRDs, UX specs, architecture, briefs). All applies require --i-approve and produce commits with Proposed-By/Approved-By trailers.

Read-only pre-flight verifier. Checks Linear, gh, omp, agent symlinks, and SpecSafe state-file integrity before a dogfood session. No mutations, no --i-approve gate.

This skill should be used when the user asks to add, update, or review evaluation coverage for agent behavior, LLM workflows, SpecSafe slices, harness prompts, or evalfly suites; when a PR changes agent instructions, slash commands, skills, schemas, or workflow-critical behavior; or when asked whether an eval is required.

Read and draft-mutate GitHub PRs, issues, and API endpoints via gh(1). All mutations require --i-approve. PR creation enforces a Linear-state invariant and is logged to .pi/.github-log.jsonl.

Fetch, cache, and expose the latest official documentation for registered libraries, SDKs, and APIs. Turns Luci's "always check latest official docs" rule from prose into an executable contract.

Read and draft-mutate Linear issues, comments, and state transitions. All mutations require --i-approve and are logged to .pi/.linear-log.jsonl.

Slice lifecycle CLI for SpecSafe — begin, end, and inspect the open slice in `.pi/.specsafe-state.json`.

Consent-gated git push with pre-flight checks and audit log. Use when pushing any branch. Dry-run by default; pass --i-approve to execute. Checks include clean tree, branch ahead of remote, not on main/master (use --allow-main to override), Spec-Slice trailer present in commit range, and no open SpecSafe slice. Appends JSON to .pi/.push-log.jsonl on success.

Remove unused code from this project with LSP-verified safety and parallel task subagents for atomic removal batches. Triggers: 'remove dead code', 'dead code', 'cleanup', 'remove unused'.

Shell-driven browser automation from bash — the `agent-browser` CLI (open/snapshot/state, `@e` refs) and the `npx playwright test` runner. Use for CI runs, headless screenshots, and test-suite execution. For in-session browsing prefer `dev-browser`.

Browser automation via the in-process Playwright/MCP-style API (the `browser` tool's `tab.*`/`page.*` helpers). Use for scripted in-session browsing, scraping, and UI testing. For general browser work prefer `dev-browser`; for shell/`npx playwright test` runs use `playwright-cli`.

Parallel security research skill. Orchestrates 3 vulnerability hunters and 2 PoC engineers via parallel task() subagent batches to audit a codebase, prove exploitability, classify root causes, and calibrate severity by actual exploitability. Use for security review, vulnerability research, exploitability audit, pre-release security check, threat model validation. Triggers: 'security-research', 'security research', 'security review', 'vulnerability audit', 'exploitability audit'.

Thorough, file-cited technical debt audit across 9 dimensions using ast_grep (tree-sitter), search, bash, and LSP. Produces TECH_DEBT_AUDIT.md with severity, effort estimates, and prioritized fixes. Use when asked for codebase health check, tech debt audit, architecture review, code quality assessment, or cleanup planning. Triggers: 'tech debt', 'technical debt', 'debt audit', 'code health', 'technical debt audit', 'codebase health check', 'find tech debt', 'debt analysis', 'audit code quality'.

Removes AI-generated code smells from a SINGLE file while preserving functionality. For multiple files, call in PARALLEL per file.

Browser automation with persistent page state. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, or automate browser workflows.

MUST USE for ANY git operations. Atomic commits, rebase/squash, history search (blame, bisect, log -S). Triggers — 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that'.

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile critics via parallel task() subagent batches for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the plan agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', 'adversarial plan', 'hostile planning', 'cross-critique plan'.

Designer-turned-developer who crafts stunning UI/UX even without design mockups. Use when implementing or polishing frontend interfaces.

Post-implementation review orchestrator. Launches parallel review sub-agents covering goal/constraint verification, code quality, security, hands-on QA, and context mining. ALL must pass for review to pass. MUST USE after completing significant implementation work.