One mission
Goal, scope, authority, constraints, proof targets, and stop conditions stay visible so parallel work does not mutate the assignment.
/suede-full-send · Intent router
Use Full Send when the outcome is broad, authorized, and consequential enough to deserve the useful lanes, hard judgment, adversarial review, and a finish backed by evidence.
Full Send is a control policy for effort. It keeps one mission record, selects one controller, opens the maximum useful set of lanes that will not collide, and requires every important completion claim to resolve to proof, an explicit gap, or a blocker.
Goal, scope, authority, constraints, proof targets, and stop conditions stay visible so parallel work does not mutate the assignment.
Parallelism earns its place through independent evidence or non-overlapping ownership. Duplicate lanes and padding do not count as effort.
The controller reconciles findings, challenges consequential calls, verifies the outcome, and reports what is proved, unproved, or blocked.
Full Send does not become a second orchestrator. It chooses the narrowest controller that can govern the mission, then stays at the intent layer.
complex
suede-agent-teams
Use communicating lanes for high-risk, cross-surface, or multi-stage work that needs shared context, review, recovery, and a signed handoff.
high-volume
suede-codex-fleet
Use independent worker units when the job benefits from volume, self-contained briefs, and mandatory controller review.
atomic
smallest specialist
Use one focused skill when the scope is narrow. Full Send does not turn a one-file fix into an orchestration ceremony.
House line
Never end your allocation above zero.
This is dry intent copy: use the available time, tools, and useful lanes to pursue the authorized outcome instead of coasting. It is not a literal counter promise, a hidden host counter, a change to model limits, or permission to spend money or act outside the user's authority.
Activity is not completion. The final mission record ties each material claim to current evidence and names anything the run could not truthfully prove.
The acceptance condition was checked against current repo state, terminal output, a live service, or another task-appropriate source of truth.
The work may exist, but the required verification was unavailable or incomplete. The handoff says exactly which proof is missing.
A real dependency, authority boundary, or external-state condition prevents completion. The blocker and smallest next action stay explicit.
The best prompt gives Full Send a real finish line, the surfaces in scope, and any hard authority boundary. It can infer the useful lanes from there.
Use $suede-full-send to finish this authorized release across code, docs, CI, and production verification. Preserve unrelated WIP and return proof.
Use $suede-full-send to audit this public launch end to end, split only non-colliding lanes, challenge the strongest claims, and fix what is in scope.
Use $suede-full-send to take this broad repo outcome to a verified handoff. No external spend or publication without my explicit approval.
Claude Code and Codex can each install the complete public pack through their native marketplace. The direct Codex installer remains available when you want only Full Send.
Claude Code full pack
/plugin marketplace add JasonColapietro/suede-creator-skills
/plugin install suede-skills@suede
Codex full pack
codex plugin marketplace add JasonColapietro/suede-creator-skills --ref main
codex plugin add suede-skills@suede-codex
Codex — Full Send only
python3 ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \
--repo JasonColapietro/suede-creator-skills \
--path skills/suede-full-send