Piagent Platform v1.51.2

Short commands, predictable behaviour

Slash commands package policy and flow, with one inspector for diffs, commands, safety and context.

Session basics

/login/model/piagent-status/piagent-inspector/piagent-orchestration/usage/usage efficiency/task-preflight/context preflight/name/new/session

Set up a session, select a model, inspect profile, usage, or orchestration state, and decide whether large work needs a new session (/new).

Project profile

/profile/profile list/profile auto/profile setup/profile tech/profile tech setup/profile fullstack/profile be-fe

Apply a profile directly or use a deterministic selector for the profile and tech stack. Output stays compact and does not require model interpretation.

Assigning work

plain languageread-only, do not editplan firstreview the diff

Since 1.9.0 there is no /workflow, /task or task contract: say what you need in plain language; the guard checks every tool call.

Git handoff

commit in plain languagedraft PR in plain languagegit statusgit diff

There is no /git namespace. Commits and PRs are ordinary requests; pushes, GitHub writes and broad staging such as git add -A require confirmation.

New sessions

/new/compact

Start a clean session when the current context is heavy or belongs to different work (/fresh was retired in 1.9.0).

MCP

/mcppiagent-mcp

Inspect and configure tool servers with the core, popular, or design presets.

Subagents

/subagents-doctor/subagents-models/subagents-fleet/subagent-cost

Inspect runtime health, model routing, background runs, and cost. Daily work remains solo-first; subagents are reserved for independent scouting, planning, or review lanes.

Product visibility

Inspect runtime decisions

These commands read deterministic state without asking a model to interpret it. They are optional in the daily workflow.

Preflight before a task
/task-preflight --json Review current repository state

Inspect intent, risk, scope, runtime, solver route, phases, tool and helper recommendations, backend, approvals, and blockers before starting.

Session status
/piagent-status

The session's profile, guard, runtime and model. It changes no file, reruns no check and starts no model turn.

Activity, diff, and context
/piagent-inspector

One menu for summary, files, commands, security, context, and toggle. A four-row panel appears automatically just above the native footer: green is healthy, yellow needs attention, and red marks failures, safety warnings, or context at 80% and above. The command reads local state without calling a model.

Sourced efficiency
/usage efficiency

Inspect context and tool utilization, verification, repair, retries, and helper usage. Runtime values that cannot be measured stay null.

Helper policy
/piagent-orchestration

Inspect the solo-first policy, read-only roles, reviewer or Oracle recommendations, owned budgets, and the one-writer invariant.

Sourced diffs, no token estimates.

The footer panel uses the files observed in the session for fast updates; the files view computes the exact snapshot delta and marks mixed-working-tree when a turn touches a file that was already dirty. Pi exposes model usage by turn and session, so per-built-in-tool tokens remain unavailable.

Optional skills and packages

Audit test quality, run experiment loops

Test audit
/skill:test-audit audit the tests of the billing module

The test-audit skill ships with Piagent: it stops new tests that protect no behavior and finds duplicated tests, tests tied to the implementation, and code that exists only for tests. The agent uses it when writing or changing tests; run the command to sweep a whole area.

Experiment loop
git switch -c experiment/test-speed

If you add a package with an automatic experiment loop (run_experiment, log_experiment): the guard checks each experiment command like any shell command, and lets the loop run only on an experiment/ branch or a worktree of its own, started with nothing uncommitted, because every dropped idea erases all uncommitted changes.