Long-Horizon Prompt Writer
Writes and checks launch prompts for long autonomous or multi-agent runs so they don't stop early or game the goal.
- Stars
- 18,282
- Licence
- MIT
- Files
- 6 files
What it does
Once installed, when you ask the agent to write or review the launch prompt for a long-running autonomous task (hours or days, possibly with many parallel sub-agents), it structures that prompt as a precise task brief. The brief defines key terms, states an exact success test, lists outcomes that do not count as success, sets rules for when to keep going or stop, and requires an adversarial check before the result is accepted. It also points you to other skills for topics like agent architecture, evaluation, and memory.
Great for
- Engineers preparing a long autonomous agent run on a hard, open-ended problem
- Teams designing an orchestrator prompt that manages many parallel worker agents
- People diagnosing a long agent run that returned early, gave a near-miss answer, or claimed completion falsely
Skip it if
- Simple, short, interactive prompts, where the extra structure adds cost without benefit (the skill itself says short prompts fail cheaply)
- Problems where a success condition can be checked by a test suite alone without extra prompt design, since the skill's own advice is to machine-check those parts with commands
- Anyone expecting it to run the long task itself: it shapes the instructions, it does not provide an execution environment or evaluator
Try asking
- “Help me turn this vague goal, 'figure out why our model drifts', into a brief with a success predicate and non-counting outcomes before I launch a 10-hour run.”
- “Review my orchestrator prompt for 20 parallel workers and tell me where workers could converge on the same approach or return fabricated completion claims.”
- “Write the output contract and acceptance checks each sub-agent must satisfy before its result is returned to me.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 6 original files, the MIT licence and an attribution note.
long-horizon-prompting.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip long-horizon-prompting.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip long-horizon-prompting.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Give the agent the full problem statement and your domain's known failure modes; the skill works best when the brief can enumerate specific ways a solution could be wrong.
- Ask the agent to separate checks a command can decide (tests, schema validation) from those that need a written success predicate.
- Read the non-counting outcomes list carefully with a domain expert, since it is the most important part for preventing near-miss answers.
What's inside
- references/cdc-prompt-annotated.md
- references/mass-handoff-contracts.md
- references/research-evidence.md
- references/task-brief-template.md
- references/vendor-guidance.md
- SKILL.md
Redistributed unmodified from muratcankoylan/Agent-Skills-for-Context-Engineering under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.