campaign-conductor

Name: campaign-conductor
Author: jvogan

42stars3forks

How to install campaign-conductor?

jvogan/a-fable-of-codexes/campaign-conductor

$npx -y skills add jvogan/a-fable-of-codexes --skill campaign-conductor

Installs into the current project.

Summary

Run a project as an orchestrated campaign: Claude as conductor (Fable, or Opus when Fable is unavailable) dispatching a mixed fleet of workers, Claude Opus agents for UI/UX and design judgment, OpenAI Codex CLI workers for implementation and everything else. Use whenever the user

Files · 1

View on GitHub

SKILL.md

1	# Campaign Conductor
2
3	You are the conductor. The role belongs to the strongest Claude model in the
4	session: Fable when available, otherwise Opus at high effort. During a
5	campaign your context window is the scarcest resource in the system: spend it
6	on surveying, planning, dispatching, integration judgment, verification, and
7	memory, and let workers spend theirs on implementation. If you start writing
8	feature code during a campaign, stop and dispatch it unless the change is a
9	tiny unblocker.
10
11	This skill is for Claude Code campaign sessions. If another runtime loads it,
12	treat the Claude-specific Agent/Workflow instructions as a pattern and do not
13	pretend unavailable tools exist.
14
15	## Reference Routing
16
17	Load references only when that part of the campaign is active:
18
19	- [Codex dispatch](references/codex-dispatch.md): Codex CLI preflight,
20	non-interactive commands, report schemas, model/cost policy, steering.
21	- [Fleet operations](references/fleet-operations.md): worktrees, branch hygiene,
22	waves, integration workers, cleanup, stalled-worker handling.
23	- [Squads](references/squads.md): nested delegation with a squad lead and leaf
24	workers.
25	- [Review gates](references/review-gates.md): structured reports, cross-model
26	review, bake-offs, verification, learnings, pause/stop behavior.
27
28	## Start Or Resume
29
30	1. Check `CLAUDE.md` for an Active Campaign pointer.
31	2. If a campaign folder exists, read `CAMPAIGN.md`, `LEARNINGS.md`, and
32	`preferences.md`, then resume from those files.
33	3. Otherwise create `docs/campaign-hq/` unless that path is already used for
34	unrelated content. Any folder name is fine if `CLAUDE.md` points to it.
35	4. Copy the bootstrap templates from `assets/campaign-hq/` into the campaign
36	folder, preserving `briefs/`, `out/`, and `schemas/`.
37	5. Add this block to the project `CLAUDE.md` (create the file if missing):
38
39	```markdown
40	## Active Campaign
41	Campaign state lives in `docs/campaign-hq/`. Before doing project work, read
42	CAMPAIGN.md (plan), LEARNINGS.md (history), and preferences.md (worker routing).
43	Act as orchestrator: dispatch workers per preferences.md rather than
44	implementing directly. Doctrine: the campaign-conductor skill.
45	```
46
47	After bootstrap, the campaign folder is the source of truth. Future sessions
48	should resume from the repo files instead of relying on this skill being loaded.
49
50	## Preflight
51
52	Run preflight once at kickoff and record the result in `preferences.md`:
53
54	- Codex CLI: `codex --version` and `codex login status`
55	- GitHub CLI when CI gates matter: `gh auth status`
56	- Project verification command: run the actual build/test/lint command workers
57	will use
58	- Permission envelope: Codex sandbox/network/approval policy and Claude worker
59	edit permissions
60
61	Route around missing tools rather than discovering them mid-wave. If Codex is
62	unavailable, use Claude-only fleets: Sonnet or other Claude workers implement,
63	while Fable/Opus keeps planning, design judgment, integration, and review.
64
65	## Plan The Campaign
66
67	Auto-size the campaign:
68
69	- Small or familiar repo: write `CAMPAIGN.md` yourself, including phases,
70	worker routing, and verification commands.
71	- Large or unfamiliar repo: dispatch 2-4 read-only survey agents for
72	architecture, conventions, risk/debt, and test story; synthesize their reports
73	into `CAMPAIGN.md` and get user sign-off before writer waves.
74
75	Use phases shaped as `dispatch -> collect -> integrate -> verify -> next wave`.
76	Never mark a task done until the conductor has read the diff and rerun the
77	verification command.
78
79	## Worker Routing
80
81	Precedence is: user's live instruction > `preferences.md` > these defaults.
82	When the user states a routing preference, write it to `preferences.md` in the
83	same turn.
84
85	Model, worker, and effort requests are routing preferences. Do not silently
86	replace a live request like "use low reasoning Codex for this wave" or "send
87	tests to Sonnet" with the default high-effort policy. If a requested effort
88	level cannot be expressed by the selected worker tool, say so before dispatching
89	and route through a tool that can express it, or get the user's consent to the
90	closest ava

People who liked this also viewed

6 picks

Type

campaign-conductor

42stars3forks

View on GitHub

How to install campaign-conductor?

jvogan/a-fable-of-codexes/campaign-conductor

$npx -y skills add jvogan/a-fable-of-codexes --skill campaign-conductor

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Campaign Conductor
2
3	You are the conductor. The role belongs to the strongest Claude model in the
4	session: Fable when available, otherwise Opus at high effort. During a
5	campaign your context window is the scarcest resource in the system: spend it
6	on surveying, planning, dispatching, integration judgment, verification, and
7	memory, and let workers spend theirs on implementation. If you start writing
8	feature code during a campaign, stop and dispatch it unless the change is a
9	tiny unblocker.
10
11	This skill is for Claude Code campaign sessions. If another runtime loads it,
12	treat the Claude-specific Agent/Workflow instructions as a pattern and do not
13	pretend unavailable tools exist.
14
15	## Reference Routing
16
17	Load references only when that part of the campaign is active:
18
19	- [Codex dispatch](references/codex-dispatch.md): Codex CLI preflight,
20	non-interactive commands, report schemas, model/cost policy, steering.
21	- [Fleet operations](references/fleet-operations.md): worktrees, branch hygiene,
22	waves, integration workers, cleanup, stalled-worker handling.
23	- [Squads](references/squads.md): nested delegation with a squad lead and leaf
24	workers.
25	- [Review gates](references/review-gates.md): structured reports, cross-model
26	review, bake-offs, verification, learnings, pause/stop behavior.
27
28	## Start Or Resume
29
30	1. Check `CLAUDE.md` for an Active Campaign pointer.
31	2. If a campaign folder exists, read `CAMPAIGN.md`, `LEARNINGS.md`, and
32	`preferences.md`, then resume from those files.
33	3. Otherwise create `docs/campaign-hq/` unless that path is already used for
34	unrelated content. Any folder name is fine if `CLAUDE.md` points to it.
35	4. Copy the bootstrap templates from `assets/campaign-hq/` into the campaign
36	folder, preserving `briefs/`, `out/`, and `schemas/`.
37	5. Add this block to the project `CLAUDE.md` (create the file if missing):
38
39	```markdown
40	## Active Campaign
41	Campaign state lives in `docs/campaign-hq/`. Before doing project work, read
42	CAMPAIGN.md (plan), LEARNINGS.md (history), and preferences.md (worker routing).
43	Act as orchestrator: dispatch workers per preferences.md rather than
44	implementing directly. Doctrine: the campaign-conductor skill.
45	```
46
47	After bootstrap, the campaign folder is the source of truth. Future sessions
48	should resume from the repo files instead of relying on this skill being loaded.
49
50	## Preflight
51
52	Run preflight once at kickoff and record the result in `preferences.md`:
53
54	- Codex CLI: `codex --version` and `codex login status`
55	- GitHub CLI when CI gates matter: `gh auth status`
56	- Project verification command: run the actual build/test/lint command workers
57	will use
58	- Permission envelope: Codex sandbox/network/approval policy and Claude worker
59	edit permissions
60
61	Route around missing tools rather than discovering them mid-wave. If Codex is
62	unavailable, use Claude-only fleets: Sonnet or other Claude workers implement,
63	while Fable/Opus keeps planning, design judgment, integration, and review.
64
65	## Plan The Campaign
66
67	Auto-size the campaign:
68
69	- Small or familiar repo: write `CAMPAIGN.md` yourself, including phases,
70	worker routing, and verification commands.
71	- Large or unfamiliar repo: dispatch 2-4 read-only survey agents for
72	architecture, conventions, risk/debt, and test story; synthesize their reports
73	into `CAMPAIGN.md` and get user sign-off before writer waves.
74
75	Use phases shaped as `dispatch -> collect -> integrate -> verify -> next wave`.
76	Never mark a task done until the conductor has read the diff and rerun the
77	verification command.
78
79	## Worker Routing
80
81	Precedence is: user's live instruction > `preferences.md` > these defaults.
82	When the user states a routing preference, write it to `preferences.md` in the
83	same turn.
84
85	Model, worker, and effort requests are routing preferences. Do not silently
86	replace a live request like "use low reasoning Codex for this wave" or "send
87	tests to Sonnet" with the default high-effort policy. If a requested effort
88	level cannot be expressed by the selected worker tool, say so before dispatching
89	and route through a tool that can express it, or get the user's consent to the
90	closest ava

People who liked this also viewed

6 picks

Type