skill-goal

Standalone Codex skill project for the goal-prompt skill.

goal-prompt turns broad work requests into durable, verifiable long-running goal prompts. It is designed for tasks that need to survive context compaction, run item-by-item for hours, handle external blockers, and close only after a current-state completion audit.

Contents

goal-prompt/
  SKILL.md
  agents/openai.yaml
  references/goal-prompt-patterns.md

agents/openai.yaml belongs inside the goal-prompt/ skill directory. It is not a repository-root agent config.

Installation / Packaging

One-line install:

tmpdir="$(mktemp -d)" && git clone --depth 1 https://github.com/kt-aicoding/skill-goal.git "$tmpdir" && mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills" && rm -rf "${CODEX_HOME:-$HOME/.codex}/skills/goal-prompt" && cp -R "$tmpdir/goal-prompt" "${CODEX_HOME:-$HOME/.codex}/skills/goal-prompt"

For local development, copy the goal-prompt/ directory into a Codex skills directory:

cp -R goal-prompt "${CODEX_HOME:-$HOME/.codex}/skills/"

The actual skill name is defined in goal-prompt/SKILL.md as goal-prompt.

Use Cases

• Convert a broad request into an active-goal objective.
• Design a long-running migration, deployment, QA, audit, or documentation goal.
• Define per-project execution loops and blocker policy.
• Write completion criteria that require real evidence rather than intent.
• Produce a short /goal command and compact create_goal objective from a full goal prompt.
• Split complex goals into a durable instruction file plus a short objective when the /goal limit would be exceeded.

Trigger Examples

Good trigger requests:

• Use $goal-prompt to design a resumable goal for migrating every repo in this workspace.
• Create a /goal objective with done criteria and blocker policy for this deployment audit.
• Turn this vague multi-project cleanup into a long-running goal prompt.
• My /goal objective exceeded 4,000 characters; rewrite it as a file-backed goal.

Non-trigger requests:

• Fix this lint error now.
• Make a short checklist.
• Brainstorm app names.
• Write generic prompt copy that is not meant to become a long-running goal.

Validation

Local structural checks:

test -f goal-prompt/SKILL.md
test -f goal-prompt/agents/openai.yaml
test -f goal-prompt/references/goal-prompt-patterns.md
python3 - <<'PY'
from pathlib import Path
for path in [
    "goal-prompt/SKILL.md",
    "goal-prompt/agents/openai.yaml",
    "goal-prompt/references/goal-prompt-patterns.md",
]:
    text = Path(path).read_text()
    assert text.strip(), path
print("basic structure ok")
PY

Optional Codex skill validation, if you have the skill-creator system skill available:

python3 "${CODEX_HOME:-$HOME/.codex}/skills/.system/skill-creator/scripts/quick_validate.py" goal-prompt

Expected output:

Skill is valid!

Safety

• Do not include real secrets, cookies, private tokens, or copied .env values in generated goals.
• Use secret names, env var names, secret-manager references, or access prerequisites instead.
• Require redacted field-level verification if local credential diagnosis is explicitly needed.
• Pause before destructive actions, paid operations, broad rewrites, or irreversible production changes unless the user already authorized them.
• Keep paste-ready /goal and create_goal objectives below platform limits; put long instructions in a referenced file.

Notes

• The repository name is skill-goal.
• The actual skill name is goal-prompt.
• Do not place secrets, private paths, tokens, or real API keys in skill files.