| name | goalify |
| description | Use this skill when the user wants to convert rough coding, product, architecture, audit, review, migration, cleanup, debugging, or orchestration intent into a compact Codex `/goal` payload. Trigger on "goalify", "Codex goal", "/goal", "make this a goal", "RALF loop", "turn this into a Codex prompt", "$goalify interactive", or requests to be questioned before creating a goal. Do not use for normal implementation or review unless the user explicitly asks to produce a Codex goal/prompt. |
| license | MIT |
| compatibility | Designed for Codex CLI 0.128+ with GPT-5.5 and the Agent Skills format. Requires the local Linux `codex-goal` helper for protected long goal files. |
| metadata | {"version":"0.2.0","author":"bengous"} |
Goalify
Convert messy intent into the smallest useful payload to put after Codex /goal.
You are not doing the downstream implementation. You are producing goal text the human can paste after /goal, or a protected local file path the human can pass to /goal.
Do not create, set, update, pause, resume, clear, or complete an actual Codex goal yourself.
Entry Gate
Before asking clarification questions or generating any goal payload, verify the mandatory helper:
test -x /usr/local/bin/codex-goal
test ! -L /usr/local/bin/codex-goal
[[ "$(/usr/bin/stat -c '%u:%g:%a' /usr/local/bin/codex-goal)" == "0:0:755" ]]
/usr/local/bin/codex-goal --version >/dev/null
If either command fails, stop. Explain briefly that codex-goal is required to write immutable goal files, especially when Codex runs with danger-full-access. Read script/references/install.md, summarize the exact system side effects, and ask for explicit approval before running any installer. Do not install automatically.
Modes
Default mode is draft-first. Use it for $goalify, make this a goal, and similar requests. Infer reasonable details and ask only when a missing detail materially changes scope, risk, or acceptance criteria.
Interactive mode is question-first. Use it when the user says $goalify interactive, interroge-moi, pose les questions une par une, build this goal with me, or equivalent. Ask one question at a time. Each question must include your recommended answer. If a question can be resolved by inspecting the codebase or local docs, inspect instead of asking.
Core Principle
/goal needs the durable destination, not a runbook.
Include only information that changes execution:
- objective
- relevant context
- constraints and side-effect boundaries
- success criteria
- validation or evidence
- stop condition
- pause/blocker condition
Do not include generic role text such as "You are Codex CLI". Do not restate obvious working-directory or repository facts unless they materially affect the goal.
Prompt Shape
Use compact GPT-5.5-oriented structure. Put the objective as the first line without an Objective: label, because Codex and Pi already wrap /goal output with their own objective label. Prefer these labels for later sections when they carry real information:
<objective sentence>
Context:
Constraints:
Success means:
Validate with:
Stop when:
Pause if:
Only include useful sections. For simple goals, fewer sections are better.
Output Rules
The human types /goal manually. Do not prefix short output with /goal.
For payloads at or under 4000 characters, output only the raw goal payload. Do not add headings, analysis, notes, context placement tables, or a separate session prompt.
For payloads over 4000 characters, write the payload automatically with codex-goal and do not reprint the payload:
/usr/local/bin/codex-goal --root "$PWD" --slug "$SLUG" <<'EOF'
<goal payload>
EOF
Use a single-quoted heredoc delimiter so shell interpolation cannot alter the payload.
After a successful write, output only the helper result shape:
Wrote: .agents/goals/<slug>.md
Protected: immutable
Usage:
/goal .agents/goals/<slug>.md
If the target root is not the current working directory, use the absolute path printed by codex-goal.
Slug Rules
Generate slugs automatically from the objective:
- lowercase ASCII
- punctuation and spaces become
-
- strict safe pattern:
[a-z0-9][a-z0-9._-]{0,80}
- no
.md suffix in the slug
- truncate to roughly 60-80 characters
- if there is a collision, add a short suffix
Ask for a slug only if one cannot be derived.
Clarification Rules
Extract only what affects the goal payload:
- task type
- deliverable
- scope boundaries
- source-of-truth files or artifacts
- acceptance criteria
- validation commands or evidence
- side-effect permissions
- manual approval gates
- stopping condition
- blockers that should pause the goal
Ask at most one question at a time in interactive mode. In default mode, ask only for blocking ambiguity.
Helper Contract
codex-goal is the required local writer for protected long goal files. Agents call /usr/local/bin/codex-goal directly; the installed wrapper performs the narrow non-interactive sudo call to the privileged helper. The helper writes under .agents/goals/, keeps .agents/ owned by the invoking user, locks only the dedicated .agents/goals/ directory and the goal file with the Linux immutable flag, keeps the goal file owned by the invoking user with mode 0444, and verifies protection before success. It must not change attributes on .codex.
Do not fall back to weak chmod-only file writing.
Quality Check
Before final output, verify:
- the output is a goal payload, not a wrapper
- short output does not start with
/goal
- short output does not start with
Objective:
- no separate session prompt exists
- no generic Codex role preamble is included
- success criteria are observable
- validation or evidence is specified when useful
- stop and pause conditions are clear
- destructive or external side effects require explicit approval
- long payloads are written through
codex-goal
