stevengonsalvez/make-a-goal
toolkit/packages/skills/make-a-goal/SKILL.md
Produce a fully-filled autonomous-run mega-prompt at .agents/goals/<slug>.md by interviewing the user for the missing context. The output file is the "set the goal · walk away · come back to shipped work" prompt — drop it into a fresh Claude Code / Codex session to run end-to-end without hand- holding. Use when Stevie says "/make-a-goal", "/make-a-goal <outcome>", "make a goal", "set a goal", "build me a goal prompt", or similar. The skill does NOT execute the goal — it only produces the artifact. Named `make-a-goal` (not `goal`) because `/goal` is a reserved system slash.
$ install --global
skillsauthnpx skillsauth add stevengonsalvez/agents-in-a-box make-a-goalInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 14, 2026, 4:36 AM141.0s3 files scanned
SKILL.md
- name:
- make-a-goal
- description:
- |
- argument-hint:
- [the final outcome — one line, optional]
/make-a-goal — autonomous-run mega-prompt builder
What it does
/make-a-goal [outcome line] interviews Stevie for the operating context
an autonomous coding agent needs, then writes a filled mega-prompt to
.agents/goals/<slug>.md. Stevie runs the goal by pasting the file
contents into a fresh session — this skill never executes the run.
The mega-prompt template lives at assets/mega-prompt.md.tmpl. The stub
plan handed to /interview lives at assets/stub-plan.md.tmpl.
Trigger
/make-a-goal— no args, interview asks for the outcome line too/make-a-goal <outcome>— outcome line pre-fills{{OUTCOME}}, interview gathers the rest- Natural language: "make a goal", "set a goal", "build me a goal prompt"
Flow
1. Capture the outcome line
If $ARGUMENTS is non-empty, treat it as the final-outcome line and
slugify it. Else, use AskUserQuestion with a single open question:
"What does done look like — in one line?"
The outcome line is the most important field. Push back on vague answers ("ship the app") and demand specificity ("deploy v1 of the dashboard at dashboard.example.com with login + 3 charts working"). One sentence, present-tense, observable.
2. Generate slug + paths
- Slug: kebab-case the outcome line, strip stopwords, truncate to 50
chars. Example: "deploy v1 of the dashboard with login" →
deploy-v1-dashboard-with-login. - Plan path:
.agents/goals/<slug>.plan.md - Spec path:
.agents/goals/<slug>.plan-spec.md(what/interviewwrites) - Goal path:
.agents/goals/<slug>.md(the final mega-prompt)
Create .agents/goals/ if missing. If <slug>.md already exists,
append a numeric suffix (<slug>-2.md).
3. Write the stub plan
Read assets/stub-plan.md.tmpl, substitute {{OUTCOME}}, and write to
the plan path. The stub is deliberately prescriptive — it tells
/interview exactly which fields to extract so the resulting spec maps
back to the mega-prompt without guesswork.
4. Delegate to /interview
Invoke the /interview skill against <slug>.plan.md. /interview
will use its mandatory structured-user-prompt tool (AskUserQuestion on
Claude Code) and produce the spec at <slug>.plan-spec.md.
Per /interview's contract, it runs iteratively until "all major areas
are covered". Trust it.
5. Map spec → mega-prompt fields, fill gaps
Read <slug>.plan-spec.md. Extract:
| Mega-prompt field | Likely spec source |
|---|---|
| {{PROJECT}} | Executive Summary · Objectives → Primary Goals |
| {{STACK}} | Technical Requirements → Components / Architecture |
| {{CURRENT_STATE}} | Implementation Notes · context paragraphs |
| {{WORKING_DIR}} | Constraints section · or not present |
| {{CONSTRAINTS}} | Constraints & Dependencies → Technical / Timeline |
| {{AUDIENCE}} | User Experience intro · or not present |
| {{SUCCESS_1..3}} | Objectives → Success Metrics |
Fields not reliably in /interview's spec template:
working dir, audience, current state. If those are missing
or thin in the spec, run one AskUserQuestion round with only the
unfilled fields. Don't re-ask anything the spec already answers.
If the spec produced more than three success metrics, ask Stevie which three are the hard-gates for "done". If it produced fewer than three, ask for the remainder — three is the minimum the template needs.
6. Render the mega-prompt
Read assets/mega-prompt.md.tmpl. Substitute every {{FIELD}} token
with the gathered value. Multi-line values: keep them single-line where
the template uses a single line (Context block) — join with ; if
needed. The Operating Rules / Quality Bar / Final Deliverable blocks
are verbatim — never edit them.
Write to <slug>.md.
7. Report back
Tell Stevie:
- where the goal file landed (absolute path)
- where the spec landed (in case he wants to revise context without re-interviewing)
- one-line instruction to run it: paste the contents of
<slug>.mdinto a fresh Claude Code or Codex session - the slug (so he can refer to it later)
Caveman mode by default. Address as Stevie.
Non-goals
- Do not execute the goal. This skill only produces the artifact.
- Do not commit
.agents/goals/. Leave that decision to Stevie — the directory may or may not be gitignored per repo. - Do not edit the Operating Rules / Quality Bar / Final Deliverable blocks in the template. They're intentionally non-negotiable across every run.
Edge cases
- No
/interviewskill available (e.g. a host without it deployed): fall back to running the question set directly viaAskUserQuestionusing the field list fromassets/stub-plan.md.tmpl. Skip the spec artifact in this mode and render straight to the mega-prompt. - Outcome line is multi-sentence: ask Stevie to compress to one line before slugging. The mega-prompt's first line is the contract.
- Existing goal with same slug: append
-2,-3, … - Working dir is "this repo": capture the absolute path of
$PWDat goal-generation time, not the literal string "this repo".
Related Skills
stevengonsalvez/reflect:cost
documentation
VerifiedTrustedCommunity
Report reflect drain spend over a time window — tokens split by cached (cache_read), uncached writes (cache_creation), and io (input+output), with a $ estimate, grouped by day / outcome / model / transcript. Reads the drainer's cost log and surfaces outlier runs and cache-reuse health (the 41.5M-token failure mode = low cache reuse + high cache writes). Use to answer "what is reflection costing me" for the last day / week.
12SKILL.mdUpdated Jun 2, 2026
stevengonsalvez/ainb-fleet:standup
development
VerifiedTrustedCommunity
Show fleet status — every claude session running on the host, merged across ainb + claude-peers broker + background jobs. Use when you need to enumerate sessions before composing an action, see which sessions have a peer registered (broker-routable) vs tmux-only, check the `summary` of each session, or pipe the list into jq for filtering. Default output: text table. Pass --format json for LLM consumption.
10SKILL.mdUpdated May 31, 2026
stevengonsalvez/ainb-fleet:sequence
testing
VerifiedTrustedCommunity
Ordered multi-step prompts to fleet targets, ack-gated between steps via JSONL assistant-turn-end detection. Use for cycles like disconnect→reconnect→verify, or any flow where step N+1 requires step N to have completed first. The skill BLOCKS until each target's transcript shows the next assistant turn finishing OR per-step timeout fires (default 300s).
10SKILL.mdUpdated May 31, 2026
stevengonsalvez/ainb-fleet:needs
development
VerifiedTrustedCommunity
Center control panel — enumerate every claude session that is blocked waiting on something: a user answer (AskUserQuestion fired), an API error retry, an idle assistant turn-end with no follow-up, or an explicit WAITING: marker. Returns rich JSON with signal kind + context per session. Use this when you've stepped away from the fleet and want one place to see everything that wants your attention and answer it.
10SKILL.mdUpdated May 31, 2026