Commands And LLM Workflows

Uses the hidden, tool-free compaction agent to produce a continuation summary. After it succeeds, future turns reuse that summary instead of the older model context. The visible transcript remains available in the current agent-shell buffer. A failed compaction leaves the previous context unchanged.

Cancels active or queued work for the current runtime session, clears its messages, tools, child jobs, approvals, selected one-shot skills, and persisted transcript, while keeping the agent-shell session handle valid. It does not accept an argument.

Inspects the relevant buffer, region, file, diff, error, or project context and explains the entry points, data flow, state changes, and ownership boundaries. It is intended for understanding existing behavior and does not make edits.

/explain why this callback can complete twice

Reconstructs the symptom, identifies the smallest plausible root cause, makes a focused change, adds or updates a regression test when practical, and runs the verification that should catch the problem.

/fix reproduce and repair the failing ledger replay test

Inspects repository documentation, manifests, tests, and workflows, then creates or refreshes the project-root AGENTS.md. Useful existing instructions are preserved and unrelated churn should be avoided.

/init include the live Emacs verification workflow

Reviews the current repository state and diff like a senior code reviewer. It prioritizes correctness bugs, regressions, unsafe edge cases, and missing tests, with findings ordered by severity. It does not edit files unless explicitly asked.

/review focus on cancellation and stale callbacks

Finds the relevant project test command, starts with the narrowest meaningful test, and broadens only when useful. The result records passed, failed, and skipped checks plus remaining risk.

/test run the focused tests for the files changed in this branch

Inspects the current Git project and writes a single org-roam note whose title is exactly the repository name. With no argument it updates * Repository Summary for the whole workspace. Trailing free-form text selects a path or semantic scope and upserts one subtree under * Scoped Summaries in the same file. Global sessions are rejected before an LLM turn is created.

The destination is magent-org-roam-directory, or org-roam-directory when the Magent option is nil. The directory must already exist. The writer owns the Org ID, repository path, analyzed Git commit, filename, and atomic update. New notes are created through the noninteractive Org-roam capture API when it is available; the canonical file-level property drawer always precedes Org keywords so Org-roam can recognize and index the node. New filenames follow YYYY-MM-DDtHHMM.org, matching the default Org-roam capture convention. Subsequent runs locate the note by its file-level REPO_PATH, so timestamped, legacy, and manually renamed files still receive single-file updates.

/summarize
/summarize only the ACP session lifecycle
/summarize lisp/magent-session.el

M-x magent-agent-shell-run-skill-command selects and runs any command-like instruction skill through the supported agent-shell frontend.

The skill-backed portion of the slash command list is extensible. Any loaded instruction skill with a default-prompt field is advertised to agent-shell as a slash command. Skills can come from bundled skills/, the user skill directories, or project-local .magent/skills/ overlays. Therefore a local installation can expose more skill commands in addition to the two fixed session controls listed above.

M-x magent-run-memory-init discovers bounded Emacs configuration sources, shows a scan plan for approval, and writes the managed profile memory file. M-x magent-run-memory-refresh repeats the scan while preserving user notes.

By default, init and refresh summarize sanitized source excerpts with the current gptel provider. Customize magent-memory-use-llm to nil to write a deterministic skeleton and source index without sending configuration excerpts to the provider.

M-x magent-run-memory-clear deactivates the managed profile and clears its generated content after confirmation. User notes and a snapshot are retained. This workflow does not perform an LLM summarization request.

M-x magent-run-doctor collects bounded evidence through trusted read-only probes, recursively redacts it, and sends one request without model tools. With a prefix argument, C-u M-x magent-run-doctor lets the user review applicable probe selection. See DOCTOR.org for the probe API and security boundary.