Bmad pulse setup
Installs and configures the PULSE module in a BMAD project. Use when the user requests 'install PULSE', 'configure PULSE', or 'setup PULSE'.From its SKILL.md
npx -y skills add nidelson/bmad-module-pulse --skill bmad-pulse-setupAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
25.7 KB, as published. Nobody here has run it
Module Setup
Overview
Installs and configures a BMad module into a project. Module identity (name, code, version) comes from ./assets/module.yaml. Collects user preferences and writes them to these files:
{project-root}/_bmad/custom/config.toml— the module section[modules.pulse]with thepulse_*values (issue #73, toml-first). This is the layerresolve_config.pyreads with higher priority than the installer defaults inconfig.toml, so the team's answers win. Existing comments and other sections in this human-owned file are preserved (tomlkit round-trip).{project-root}/_bmad/config.yaml— shared project config: core settings at root only (e.g.output_folder,document_output_language). The legacypulse:module section is no longer written here; a stale one is stripped on run (that strip is the yaml→toml migration). User-only keys (user_name,communication_language) are never written here.{project-root}/_bmad/config.user.yaml— personal settings intended to be gitignored:user_name,communication_language, and any module variable markeduser_setting: truein./assets/module.yaml. These values live exclusively here.{project-root}/_bmad/module-help.csv— registers module capabilities for the help system.
The config scripts use an anti-zombie pattern — the module's answered keys are overwritten with fresh values on each run, and the stale legacy pulse: yaml section is removed, so stale values never persist.
Migrating an existing yaml install (issue #73): re-running this setup is the migration path. Before prompting, read the current effective values (via resolve_config.py --key modules.pulse and the legacy pulse: section of config.yaml) and offer them as the prompt defaults, so accepting the defaults carries the team's existing answers into custom/config.toml and strips the legacy yaml section.
{project-root} is a literal token in config values — never substitute it with an actual path. It signals to the consuming LLM that the value is relative to the project root, not the skill root.
On Activation
- Read
./assets/module.yamlfor module metadata and variable definitions (thecodefield is the module identifier) - Check if
{project-root}/_bmad/config.yamlexists — if a section matching the module's code is already present, inform the user this is an update - Check for per-module configuration at
{project-root}/_bmad/pulse/config.yamland{project-root}/_bmad/core/config.yaml. If either file exists:- If
{project-root}/_bmad/config.yamldoes not yet have a section for this module: this is a fresh install. Inform the user that installer config was detected and values will be consolidated into the new format. - If
{project-root}/_bmad/config.yamlalready has a section for this module: this is a legacy migration. Inform the user that legacy per-module config was found alongside existing config, and legacy values will be used as fallback defaults. - In both cases, per-module config files and directories will be cleaned up after setup.
- If
If the user provides arguments (e.g. accept all defaults, --headless, or inline values like user name is BMad, I speak Swahili), map any provided values to config keys, use defaults for the rest, and skip interactive prompting. Still display the full confirmation summary at the end.
Reconcile Installed Skills (Self-Heal)
Run this first, before any other step. The upstream BMAD installer
performs an additive deploy when PULSE is already installed: brand-new
files are copied, but pre-existing files (module.yaml, SKILL.md, …) are
never overwritten and renamed/removed skills are never pruned. Updating
over an existing install therefore leaves a mixed state stuck at the old
version (see issue #36). This step force-syncs every PULSE skill directory
from the authoritative source (the freshly-fetched BMAD custom-module
cache) and prunes orphaned renamed folders.
python3 ./scripts/reconcile-skills.py --project-root "{project-root}"
The script is idempotent — on an already-in-sync tree it writes nothing and
reports action: up_to_date. It is non-fatal on fresh installs: when no
source/cache is found it exits 0 with action: skipped_no_source, so
continue normally. Exit codes: 0=success, 1=validation error, 2=runtime
error. Surface any non-zero exit and stop.
Inspect the JSON action field:
up_to_date/skipped_no_source— continue silently.reconciled— reportfrom_version → to_versionto the user. If the payload includes anoticesayingbmad-pulse-setupwas updated in place, tell the user to re-run/bmad-pulse-setuponce more so the current version of this skill executes, then stop.
Run ./scripts/reconcile-skills.py --help for full usage (including
--source and --dry-run).
Collect Configuration
Ask the user for values. Show defaults in brackets. Present all values together so the user can respond once with only the values they want to change (e.g. "change language to Swahili, rest are fine"). Never tell the user to "press enter" or "leave blank" — in a chat interface they must type something to respond.
Default priority (highest wins): existing new config values > legacy config values > ./assets/module.yaml defaults. When legacy configs exist, read them and use matching values as defaults instead of module.yaml defaults. Only keys that match the current schema are carried forward — changed or removed keys are ignored.
Core config (only if no core keys exist yet): user_name (default: BMad), communication_language and document_output_language (default: English — ask as a single language question, both keys get the same answer), output_folder (default: {project-root}/_bmad-output). Of these, user_name and communication_language are written exclusively to config.user.yaml. The rest go to config.yaml at root and are shared across all modules.
Module config: Read each variable in ./assets/module.yaml that has a prompt field. Ask using that prompt, pre-filling the default with the current effective value when re-running on an existing install: prefer the resolved modules.pulse value (resolve_config.py --key modules.pulse), then the legacy pulse: value in config.yaml, then any legacy per-module value, then the module.yaml default. This is what carries an existing team's answers into custom/config.toml on the migration re-run.
Validation rules:
- If
pulse_estimation_method=story_pointsandpulse_story_point_hours_factorhas not been set, warn before continuing - If
pulse_estimation_method=bcp: do not requirepulse_story_point_hours_factor(no factor applies —estimated_hourscomes from a complexity score times the category baseline, not from converting a point value). Scoring lives in this module, in thebmad-bcp-*skills, so no companion install is required. Two further setup steps apply and are described below: the--with-bcpcustomize variant, and giving thebcp_*settings a home. - If
pulse_dev_categories=custom, request a comma-separated list
Write Files
Write a temp JSON file with the collected answers structured as {"core": {...}, "module": {...}} (omit core if it already exists). Then run both scripts — they can run in parallel since they write to different files:
uv run ./scripts/merge-config.py --config-path "{project-root}/_bmad/config.yaml" --user-config-path "{project-root}/_bmad/config.user.yaml" --module-yaml ./assets/module.yaml --answers {temp-file} --legacy-dir "{project-root}/_bmad"
python3 ./scripts/merge-help-csv.py --target "{project-root}/_bmad/module-help.csv" --source ./assets/module-help.csv --legacy-dir "{project-root}/_bmad" --module-code pulse
Note (PEP 723): only
merge-config.pyruns viauv run— it declares third-party dependencies (pyyaml,tomlkit) in its inline# /// scriptheader. Plainpython3fails withError: tomlkit is required (PEP 723 dependency).uv runreads the header and provisions the libraries in an ephemeral environment. The other scripts in this skill are stdlib-only (dependencies = []) and run under plainpython3.
merge-config.py writes the [modules.pulse] section to {project-root}/_bmad/custom/config.toml (derived from the --config-path directory; override with --custom-config-path), writes core keys to config.yaml, strips any legacy pulse: yaml section, and writes user settings to config.user.yaml. Check custom_config_path and module_keys in the output.
Both scripts output JSON to stdout with results. If either exits non-zero, surface the error and stop. The scripts automatically read legacy config values as fallback defaults, then delete the legacy files after a successful merge. Check legacy_configs_deleted and legacy_csvs_deleted in the output to confirm cleanup.
Run uv run ./scripts/merge-config.py --help or ./scripts/merge-help-csv.py --help for full usage.
Register Agent in Party Mode Roster
After writing config and help CSV, register the PULSE agent (Maxine) so it joins the Party Mode roster and other agent-aware features.
Party Mode builds its roster from the [agents] table resolved by resolve_config.py — a deep-merge of {project-root}/_bmad/config.toml (base) and {project-root}/_bmad/custom/config.toml (team). Official modules get their [agents.*] entries written into the base config.toml by the BMAD core installer, but a custom module like PULSE is never written there — so without this step the agent is installed as a skill yet stays invisible to Party Mode (/bmad-party-mode never lists her). Register her in the team-owned custom/config.toml layer, which survives re-install:
uv run ./scripts/register-party-agent.py --project-root "{project-root}" --fragment ./assets/agent-manifest-fragment.csv
The script upserts [agents.bmad-agent-pulse] (anti-zombie, idempotent) from the agent-manifest-fragment.csv values, preserving existing comments and sections (tomlkit round-trip). It runs via uv run for its PEP 723 tomlkit dependency. Check agent_key and custom_config_path in the JSON output.
If successful, inform the user: "Agent Maxine registered in the Party Mode roster (_bmad/custom/config.toml -> [agents.bmad-agent-pulse]) — run /bmad-party-mode to see her. To spotlight her, add a curated party group (e.g. a delivery/retro room) to _bmad/custom/bmad-party-mode.toml."
When the script reports "action": "migrated", also tell the user which fields it corrected and why: migrated_fields lists entries that still held a value a previous release of PULSE wrote (e.g. the retired name Levi), so refreshing them restores the truth rather than overwriting a choice they made.
Legacy agent-manifest.csv (optional, compat)
Older BMAD layouts also read {project-root}/_bmad/_config/agent-manifest.csv. If that file exists, additionally merge the fragment into it (harmless where unused). If it does not exist, skip — the [agents] registration above is what Party Mode actually reads.
python3 ./scripts/merge-help-csv.py --target "{project-root}/_bmad/_config/agent-manifest.csv" --source ./assets/agent-manifest-fragment.csv --module-code pulse
Create Output Directories
After writing config, create any output directories that were configured. For filesystem operations only (such as creating directories), resolve the {project-root} token to the actual project root and create each path-type value from config.yaml that does not yet exist — this includes output_folder and any module variable whose value starts with {project-root}/. The paths stored in the config files must continue to use the literal {project-root} token; only the directories on disk should use the resolved paths. Use mkdir -p or equivalent to create the full path.
Cleanup Legacy Directories
After both merge scripts complete successfully, remove the installer's package directories. Skills and agents in these directories are already installed at .claude/skills/ — the _bmad/ directory should only contain config files.
python3 ./scripts/cleanup-legacy.py --bmad-dir "{project-root}/_bmad" --module-code pulse --skills-dir "{project-root}/.claude/skills"
The script verifies that every skill in the legacy directories exists at .claude/skills/ before removing anything. Directories without skills (like _config/) are removed directly. If the script exits non-zero, surface the error and stop. Missing directories (already cleaned by a prior run) are not errors — the script is idempotent.
Check directories_removed and files_removed_count in the JSON output for the confirmation step. Run ./scripts/cleanup-legacy.py --help for full usage.
Generate Customize Overrides (Auto-Tracking)
After cleanup, configure auto-tracking by emitting two customize.toml
overrides in the consumer project. Auto-tracking now uses BMAD v6.4.0's
TOML-based customization framework instead of injecting steps into
workflow.md. This survives BMAD core upgrades because the override files
live in _bmad/custom/, which BMAD never overwrites.
Capability Gate
Run the capability detector first:
python3 ./scripts/detect_bmad_capability.py --project-root "{project-root}"
The script exits 0 (BMAD supports TOML customization), 1 (BMAD ≤6.3.x), or 2 (BMAD not installed) and prints a JSON payload to stdout describing the detection.
- Exit 0 — proceed with override emission below. Read
capabilityandinject_targetsfrom the payload — do not assume the targets. Exit 0 covers two different architectures:"capability": "bmad-build"— unified architecture,inject_targetsis["bmad-build"]"capability": "bmad-6.4.0+"— split architecture,inject_targetsis["bmad-dev-story", "bmad-code-review"]
- Exit 1 — abort with this message: "PULSE v0.4.0 requires BMAD ≥6.4.0.
Detected BMAD ≤6.3.x. Either upgrade BMAD (
npx bmad-method install) or pin to PULSE v0.3.x via--version." - Exit 2 — abort: "BMAD is not installed in this project root. Run
npx bmad-method installfirst, then re-run/bmad-pulse-setup."
Why the payload and not the exit code. BMAD keeps
bmad-dev-storyon disk as a deprecated shim after the unified architecture lands, so both layouts look alike from the exit code alone. Injecting the split pair on abmad-buildproject writes the hooks into workflows the user never invokes: the gate passes, this skill reports success, and auto-tracking never fires. The absence ofpulse_metricsweeks later is the only symptom.
Cleanup Legacy Markers
Even on a fresh BMAD ≥6.4.0 install, scan for stale <!-- PULSE:auto-inject -->
blocks in workflow.md left over from previous PULSE versions. The script
silently skips when workflow.md is absent (BMAD 6.4.0 removed it from
bmad-dev-story).
python3 ./scripts/cleanup-legacy.py \
--remove-pulse-markers \
--project-root "{project-root}"
The script exits 0 in all normal cases and prints {"removed": <int>, "path": "..."}
on stdout (or {"removed": 0, "skipped": "workflow.md not found", "path": "..."} when
the file is absent — this is the common case on fresh BMAD 6.4.0 installs and is NOT
a failure). A non-zero exit indicates --project-root is missing or the file system
is broken; surface the error and stop.
Cleanup Legacy bmad-pulse-agent-levi/ Folder
Pre-v0.4.5 PULSE distributed its own bmad-pulse-agent-levi/ skill folder
in parallel with the canonical bmad-agent-pulse/ folder auto-provisioned
by the BMAD installer from agent-manifest-fragment.csv. On BMAD v6.6.0
projects this produced two divergent entry points for the same Levi agent.
This step removes the legacy folder when the canonical folder is present. The script refuses to remove the legacy folder if the canonical replacement is missing, so a partial-state install is never stranded without an agent.
python3 ./scripts/cleanup-legacy.py \
--remove-legacy-agent \
--project-root "{project-root}"
The script exits 0 in every non-fatal case and prints a JSON payload with
an action field: removed, skipped_already_absent, or
skipped_no_canonical. Surface the notice to the user when action
is removed so they understand which folder was deleted. A non-zero exit
indicates --project-root is missing or the file system is broken;
surface the error and stop.
Emit Override Files
Emit one override file per entry in inject_targets from the capability gate
above — one file on the unified architecture, two on the split one. The
conflict policy is abort + --force: if a destination already exists, the
script exits 3 and the file is left untouched (sha256-stable). The user can
re-run with --force after inspecting the conflict.
Unified architecture ("capability": "bmad-build") — a single file carries
both hooks, because implementation and review run in one workflow:
python3 ./scripts/inject_customize.py \
--project-root "{project-root}" \
--skill bmad-build
Split architecture ("capability": "bmad-6.4.0+") — two files, because
bmad-dev-story ends at status "review" and only bmad-code-review reaches
"done":
python3 ./scripts/inject_customize.py \
--project-root "{project-root}" \
--skill bmad-dev-story
python3 ./scripts/inject_customize.py \
--project-root "{project-root}" \
--skill bmad-code-review
Do not emit both sets. On a bmad-build project the split pair is inert, and
bmad-code-review would fire track-done a second time if the user ever invoked
that skill directly.
BCP variant — only when scoring is enabled
If, and only if, the resolved pulse_estimation_method is bcp, append
--with-bcp to the invocation for bmad-build (unified) or
bmad-code-review (split):
python3 ./scripts/inject_customize.py \
--project-root "{project-root}" \
--skill bmad-build --with-bcp
The variant writes to the same destination — it replaces the plain template
rather than adding a second file. Its on_complete carries both steps in one
authored sequence: track-done first, then BCP recalibrate, which reads the
actual_hours that only exists once track-done has finished.
--with-bcp is rejected for bmad-dev-story (exit 2): that file carries
track-start, which recalibration does not extend.
Then emit the scoring trigger, which is not tier-dependent — the same file
on both architectures, and not listed in inject_targets:
python3 ./scripts/inject_customize.py \
--project-root "{project-root}" \
--skill bmad-create-story --with-bcp
This one closes the loop. The templates above run after implementation —
track-done, then recalibrate — and recalibrate checks for bcp.total before
doing anything. Without a hook that produces bcp.total in the first place, it
finds nothing and skips, silently and by design, forever.
bmad-create-story is a core BMAD skill under both architectures. bmad-build
does not replace it: bmad-build writes a spec, whose frontmatter carries
title, type and status but no estimated_hours to derive; it consumes a
story only by story_key. So story authoring is where scoring belongs, and it
is the same place either way.
bmad-create-story has no plain template — emitting it without --with-bcp
exits 2. PULSE has nothing to say to story authoring unless scoring is on.
Give the bcp_* settings a home that survives
Still only when pulse_estimation_method is bcp. The scoring skills read
their settings — the baseline seed, the confidence threshold, the frozen
reference rate — through bcp_config.py. Resolve what a project has now:
python3 {project-root}/.claude/skills/bmad-bcp-score/scripts/bcp_config.py \
--project-root "{project-root}"
The sources map in that output names the layer each value came from. Act on
what it says:
- Any key reads
modules.bcp. The project still has the standalonebmad-module-bcpinstalled, and that table is written by the installer from that module'smodule.yaml. Uninstalling it — the end state of issue #84 — deletes the table, and resolution falls through to the built-in defaults without failing. Copy those values into_bmad/custom/config.tomlunder[modules.pulse]now, while the source is still readable. Show the user the block before writing it. - Any key reads
default. Nothing is configured. The defaults are a floor chosen so the module runs, not an estimate of this team. Tell the user which keys are unset and what each one does;bcp_baseline_seedandbcp_reference_h_per_bcpare the two worth deciding deliberately, since one seeds every cold-start estimate and the other is the frozen denominator the leverage-vs-reference figure is measured against. Write the agreed values to the same place. - Keys read
modules.pulse. Already home. Nothing to do.
Write to _bmad/custom/config.toml, not _bmad/config.toml: the installer
regenerates the latter on every run and would drop the block. Merge into an
existing [modules.pulse] table rather than appending a second one.
Why these are not install prompts. They were, in the standalone module — installing it was itself the opt-in, so asking ten questions was reasonable. PULSE installs for everyone, most of whom estimate in hours, and BMAD's
module.yamlhas no conditional prompts. Ten questions that do not apply is a worse failure than a setup step that only runs when scoring is on.
Create the baseline
Still only when pulse_estimation_method is bcp. Every scoring run reads
bcp-baseline.yaml, and apply_score.py fails outright when it is absent — so
a project that enables scoring without this file has six skills that cannot run.
Create it with the values resolved in the previous step:
python3 ./scripts/seed_baseline.py \
--baseline-path "{project-root}/<resolved bcp_baseline_path>" \
--seed <bcp_baseline_seed> \
--min-samples <bcp_baseline_min_samples> \
--rolling-window <bcp_baseline_rolling_window>
Idempotent: an existing baseline is left untouched and the script reports
action: skipped_exists. That matters for a project migrating off the
standalone module — its baseline already holds real calibration history, and
overwriting it would discard every sample the team accumulated. Do not pass
--force on a migration; only when the user explicitly asks to start over.
The file records a config_snapshot of the three values above alongside an
empty categories map. Categories fill in later, from real actual_hours, via
bmad-bcp-recalibrate — until a category reaches min_samples it stays on the
seed and is marked is_seed: true.
Why the recalibrate step is not simply always present. It could be written to check for
bcp.totaland skip when absent — it does exactly that when present. But an instruction the agent reads on every single run, in a project that will never satisfy it, is not free: it is context spent, and it is one more thing implying the project is missing something. With scoring off, PULSE is the baseline product, not a degraded one. The opt-in lives in which file gets written, not in a runtime check.
If either invocation exits 3, surface the message to the user verbatim
(it includes the destination path and instructs how to re-run with
--force). Do NOT auto-retry with --force — the choice is the user's.
.gitignore Allowlist Snippet
Print a copy-paste-ready snippet for the consumer's .gitignore so that
_bmad/custom/*.toml is committed (team overrides) while *.user.toml
stays private. The script is read-only — never modifies the file.
python3 ./scripts/print_gitignore_snippet.py --project-root "{project-root}"
Surface the script's stdout to the user.
Post-Injection
Inform the user:
Name the files you actually wrote and the workflow the user actually runs —
a message naming /bmad-dev-story on a bmad-build project is how a broken
integration passes for a working one.
- Unified architecture: "PULSE auto-tracking integrated via
_bmad/custom/bmad-build.toml. Every story will now automatically track start and completion during/bmad-build." - Split architecture: "PULSE auto-tracking integrated via
_bmad/custom/bmad-dev-story.tomland_bmad/custom/bmad-code-review.toml. Every story will now automatically track start (during/bmad-dev-story) and completion (after/bmad-code-review)." - "To disable: delete the emitted
.tomlfile(s) from_bmad/custom/." - "To customize: edit the files manually. Re-running
/bmad-pulse-setupwill abort if you changed them — pass--forceonly if you want PULSE's defaults restored."
If any step above failed, do NOT block the rest of the setup. Report the failure clearly and continue — the user can rerun the failing piece later.
Confirm
Use the script JSON output to display what was written — config values set (written to config.yaml at root for core, module section for module values), user settings written to config.user.yaml (user_keys in result), help entries added, fresh install vs update. If legacy files were deleted, mention the migration. If legacy directories were removed, report the count and list (e.g. "Cleaned up 106 installer package files from pulse/, core/ — skills are installed at .claude/skills/"). Then display the module_greeting from ./assets/module.yaml to the user.
Outcome
Once the user's user_name and communication_language are known (from collected input, arguments, or existing config), use them consistently for the remainder of the session: address the user by their configured name and communicate in their configured communication_language.