# Multica Goal Tracker

> |

- **Type:** Skill
- **Install:** `agentstack add skill-miaodx-intuitive-flow-multica-goal-tracker`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [MiaoDX](https://agentstack.voostack.com/s/miaodx)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [MiaoDX](https://github.com/MiaoDX)
- **Source:** https://github.com/MiaoDX/intuitive-flow/tree/main/skills/multica-goal-tracker

## Install

```sh
agentstack add skill-miaodx-intuitive-flow-multica-goal-tracker
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Multica Goal Tracker

Use this skill to keep Multica issues consistent when the human workflow is:

1. Create or reuse a Multica issue.
2. Paste the `/goal` command that will drive a long `$intuitive-flow` or
   related skill run.
3. After the run completes, attach concise completion evidence.
4. If the result is incomplete, keep the issue open and append the next goal as
   another attempt on the same issue.
5. Repeat until the issue-level outcome is complete. Finish comments show the
   per-goal attempt plus cumulative issue time.

The skill can also create the Multica issue from an approved
`$intuitive-preflight` contract. That path records planning provenance and the
exact executable `/goal`, then immediately appends the normalized tracker start
comment. It does not execute the goal and does not create finish evidence.

Prefer appending comments over rewriting issue descriptions. Only update the
description when the user explicitly wants the issue top-level description
cleaned up.

All comments created by this skill use the current Multica user token, but they
must begin with the visible marker:

```markdown
> Agent 提交：以下内容由 Agent 帮忙整理并提交，用于和人工手写评论区分。
```

Keep this marker as the first visible line on start, finish, and final-review
comments.

## Create From Preflight

Use this after an `$intuitive-preflight` contract is approved and before the
goal is executed. The preflight must be `DRAFT`/approved, not
`BLOCKED_NEEDS_DECISION`, and it must contain either `Main-session /goal prompt`
or `To execute` with the executable `/goal`.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  create-from-preflight \
  --preflight-file /tmp/preflight.md \
  --workspace-id roboclaws
```

Never rely on the Multica CLI default workspace when creating an issue from a
preflight. The default workspace may point at another product. If the user gives
a URL such as `https://multica.evad.mioffice.cn/roboclaws/`, pass
`--workspace-id roboclaws` or the full URL so the script can resolve and verify
the real workspace id before writing.

Keep `create-from-preflight` issue descriptions intentionally short. Do not
paste the full `## Preflight Contract` into the issue body; the plan file is
the source of truth for detailed scope, non-goals, context, and verification.
Humans should be able to scan the issue quickly, then open the plan only when
they need the full contract.

Use `--dry-run` first when validating a new preflight shape. Run
`bun skills/multica-goal-tracker/scripts/track_goal.ts --help` for the current
option surface.

Do not use `create-from-preflight` for conversation-only work unless the
preflight body contains the full approved contract; otherwise context
compression can erase the issue's source of truth. Do not treat the created
issue or start comment as completion evidence. Finish evidence still must come
from a real run/session as described below.

## Start Tracking

Use this after the issue exists and contains the goal, or when the user gives
you the goal text directly. Run the command from the Intuitive Flow checkout
root so the script path stays portable across clone locations.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  start \
  --issue MIA-40
```

If the goal is not already in the issue description, pass `--goal-file`.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  start \
  --issue MIA-40 \
  --goal-file /tmp/goal.txt
```

Pass `--update-description` only when the user asks to normalize the issue
description itself. The script inserts or replaces a marked summary block and
preserves the existing description below it.

Use `--dry-run` before writing to a real issue when testing a prompt or script
change.

## Finish Tracking

Use this after the goal run is complete or after the user asks to preserve the
  state of a goal attempt on an existing issue. Do not create a new issue for this
flow unless the user explicitly asks. Finish evidence must come from real
session output: Multica execution run messages by default, a Codex JSONL
session, a skill-runner run directory, or an explicit transcript via
`--session-file`.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  finish \
  --issue MIA-40
```

The script picks the active goal, extracts real run/session evidence, records
attempt metadata, and posts one scannable comment with overview, timeline,
details, and raw selected output. `complete` attempts are completion records;
`partial`, `blocked`, and `failed` attempts are execution records.

If the issue has no Multica run history, finish fails fast instead of creating a
fake proof card. In that case pass a real Codex session JSONL for the finished
attempt:

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  finish \
  --issue MIA-40 \
  --session-file ~/.codex/sessions/2026/06/04/rollout-....jsonl
```

Or pass a real skill-runner run directory. It uses compact artifacts and avoids
noisy `terminal.log` output.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  finish \
  --issue MIA-40 \
  --session-dir ~/.cache/skill-runner/runs/
```

Use `--allow-manual-summary --summary "..."` only when the user explicitly
accepts a manual fallback. Manual fallback is not real session proof.

If an attempt is incomplete, finish it with `--attempt-status partial|blocked|failed`,
then run `start` again for the follow-up goal. The next finish becomes the next
attempt and the card shows cumulative issue time.

## Final Review

Use this when the human wants one final review thread for an issue that already
has multiple goal attempts, especially when an earlier attempt was partial and a
follow-up goal completed the issue. Do not hand-compose the comment. Put the
attempt data in JSON and let the script own the format.

```bash
bun skills/multica-goal-tracker/scripts/track_goal.ts \
  final-review \
  --issue MIA-40 \
  --attempts-file /tmp/multica-goal-attempts.json
```

`final-review` posts one text comment with a short issue-level summary, compact
attempt list, overview, timeline, details, and complete raw outputs for each
attempt. The comment stores metadata for every attempt, so later tracker runs
can recover cumulative duration even if older Agent comments are cleaned up.

## Useful Options

Use `--help` for full syntax. The high-value knobs are `--workspace-id`,
`--goal-file`, `--session-file`, `--session-dir`, `--run-id`,
`--attempt-status`, `--proof`, `--dry-run`, and `--allow-manual-summary`.

## Style

Default generated issue text to Chinese. Keep summaries short. The issue should
answer "what is this goal trying to accomplish?" without becoming a plan
document. Preserve the raw `/goal` block because it is the execution provenance.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [MiaoDX](https://github.com/MiaoDX)
- **Source:** [MiaoDX/intuitive-flow](https://github.com/MiaoDX/intuitive-flow)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-miaodx-intuitive-flow-multica-goal-tracker
- Seller: https://agentstack.voostack.com/s/miaodx
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.