Forward logo

Forward

Organization
Soul-Brews-Studio
forward

Hand off the current session to the next one. Use when user says "forward", "handoff", "wrap up", or before ending a session.

Overview

PublisherSoul-Brews-Studio
Repositoryarra-oracle-skills-cli
Skill nameforward
Stars
125
Forks
59
Bundled files
1
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 1 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by Soul-Brews-Studio on GitHub. Read the source before you install it.

Installation

Install the Forward AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/Soul-Brews-Studio/arra-oracle-skills-cli/tree/alpha/skills/forward
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/Soul-Brews-Studio/arra-oracle-skills-cli.git /tmp/arra-oracle-skills-cli
mkdir -p .claude/skills
cp -r /tmp/arra-oracle-skills-cli/skills/forward .claude/skills/forward
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Forward in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Forward on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Forward is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

/forward - Handoff to Next Session

Create context for next session, then enter plan mode to define next steps.

Usage

/forward              # Create handoff, show plan, wait for approval
/forward --bg         # Hand the whole handoff to a background Haiku — returns instantly
/forward asap         # Create handoff + commit immediately (no approval needed)
/forward --only       # Create handoff only, skip plan mode

/forward --bg

Fire-and-forget. A background Haiku subagent mines the session JSONL and writes the handoff file. Returns immediately — no plan mode, no blocking. Use it to snapshot progress mid-session, or at the end when you'd rather keep moving. For the full interactive flow with plan approval, use plain /forward.

Haiku is the right tier here: this is bullet extraction from structured JSONL, not writing. (Its sibling /rrr --bg uses Sonnet — a retro's diary and lessons are writing.)

Resolve first, then spawn

bash
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
ENCODED_PWD=$(echo "$ORACLE_ROOT" | sed 's|^/|-|; s|[/.]|-|g')
PROJECT_DIR="$HOME/.claude/projects/${ENCODED_PWD}"
LATEST_JSONL=$(ls -t "$PROJECT_DIR"/*.jsonl 2>/dev/null | head -1)
SESSION_ID=$(basename "$LATEST_JSONL" .jsonl | cut -c1-8)
PSI=$(readlink -f "$ORACLE_ROOT/ψ" 2>/dev/null || echo "$ORACLE_ROOT/ψ")
REPO_NAME=$(basename "$ORACLE_ROOT")
DATE_SLUG=$(date +%Y-%m-%d_%H-%M)

Substitute every <PLACEHOLDER> with the real value before spawning — the subagent cannot see your shell. Check whether the handoff directory exists; if not, tell the subagent to mkdir -p it.

Spawn ONE background Agent (model: "haiku", run_in_background: true)

You are a session archivist. Write a handoff file for the current session.

JSONL path: <LATEST_JSONL>
Vault path (PSI): <PSI>
Session ID: <SESSION_ID>
Repo: <REPO_NAME>
Date: <YYYY-MM-DD HH:MM>

Steps:
1. Run: python3 "$(ls ~/.claude/skills/forward/scripts/dig-session.py \
             ~/.claude/plugins/marketplaces/*/skills/forward/scripts/dig-session.py \
             ~/.claude/plugins/cache/*/*/*/skills/forward/scripts/dig-session.py \
             ~/.claude/skills/forward/dig-session.py 2>/dev/null | head -1)" "<LATEST_JSONL>"
2. From the JSON output, write 4-8 bullets summarizing what happened. Be specific:
   skills invoked, teammates used, files changed, key outputs.
3. Run: git -C <ORACLE_ROOT> status --short
4. Write the handoff to: <PSI>/inbox/handoff/<DATE_SLUG>_bg-forward.md

Format:

# Handoff: [session focus from the JSONL]

**Date**: <YYYY-MM-DD HH:MM>
**Source**: /forward --bg (JSONL-mined, not memory)
📡 Session: <SESSION_ID> | <REPO_NAME>

## What We Did
<bullets from the JSONL analysis>

## Uncommitted Files
<from git status --short, or "none">

## Pending
<infer from the session: what was in progress? what was mentioned but not done?>

## Next Session
<1-3 specific actions based on where the session ended>

After writing, print: "📤 Handoff written: <absolute_path>"
Do NOT enter plan mode. Do NOT commit anything.

After spawning

Tell the user where it will land, then keep working — don't wait:

📤 /forward --bg launched — handoff writing in background.
   Session: <SESSION_ID>
   Will write to: <PSI>/inbox/handoff/<DATE_SLUG>_bg-forward.md

Sibling: /rrr --bg (retrospective). Run both at session end — /rrr --bg looks back, /forward --bg looks forward.

Steps

  1. Git status: Check uncommitted work
  2. Detect session: Current session ID for traceability
  3. Session summary: What we did (from memory)
  4. Pending items: What's left
  5. Next steps: Specific actions

Session Detection

bash
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || pwd)
ENCODED_PWD=$(echo "$ORACLE_ROOT" | sed 's|^/|-|; s|[/.]|-|g')
PROJECT_DIR="$HOME/.claude/projects/${ENCODED_PWD}"
LATEST_JSONL=$(ls -t "$PROJECT_DIR"/*.jsonl 2>/dev/null | head -1)
if [ -n "$LATEST_JSONL" ]; then
  SESSION_ID=$(basename "$LATEST_JSONL" .jsonl)
  echo "SESSION: ${SESSION_ID:0:8}"
fi

Include in handoff header if detected:

markdown
📡 Session: 74c32f34 | repo-name | Xh XXm

Skip silently if detection fails.

Output

Resolve vault path first:

bash
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] || [ -L "$ORACLE_ROOT/ψ" ]; }; then
  PSI=$(readlink -f "$ORACLE_ROOT/ψ" 2>/dev/null || echo "$ORACLE_ROOT/ψ")
else
  PSI=$(readlink -f ψ 2>/dev/null || echo "ψ")
fi

Write to: $PSI/inbox/handoff/YYYY-MM-DD_HH-MM_slug.md

IMPORTANT: Always use the resolved $PSI path, never the ψ/ symlink directly. This ensures handoffs go to the project's vault (wherever ψ points). Do NOT git add vault files — they are shared state, not committed to repos.

markdown
# Handoff: [Session Focus]

**Date**: YYYY-MM-DD HH:MM
**Context**: [%]

## What We Did
- [Accomplishment 1]
- [Accomplishment 2]

## Pending
- [ ] Item 1
- [ ] Item 2

## Next Session
- [ ] Specific action 1
- [ ] Specific action 2

## Key Files
- [Important file 1]
- [Important file 2]

Confirm handoff write (announce-mode — absolute paths required)

announce-mode → absolute path (no ψ/, no ~/, no $VAR, no ...).

Use: echo "marker: $RESOLVED_PATH" — bash substitutes. See CONVENTIONS.md.

bash
HANDOFF_FILE="$PSI/inbox/handoff/$(date +%Y-%m-%d_%H-%M)_${SLUG}.md"
echo "📤 Handoff: $HANDOFF_FILE"

Then: Create Issues from Pending Items

After writing the handoff file, extract actionable items and offer to create GitHub issues.

Step 1: Extract Items

From the handoff you just wrote, collect all - [ ] items from Pending and Next Session sections.

Step 2: Filter Actionable Items

Skip items that are NOT actionable:

  • Items containing "monitor", "watch", "track", "deferred", "maybe", "consider"
  • Items that are vague (less than 4 words after the checkbox)

Step 3: Check for Duplicates

bash
# For each item, check if an issue already exists with a similar title
gh issue list --state open --search "ITEM_TITLE" --json title --jq '.[].title' 2>/dev/null

Skip items that already have a matching open issue (case-insensitive title match).

Step 4: Show and Confirm

Display the list of new issues to create:

📋 Create GitHub issues from pending items?

  1. Fix awaken git push auth
  2. /rrr --deep time-based

Create these 2 issues? [y/N]

NEVER auto-create issues without user approval.

If user declines, skip issue creation and continue to plan mode.

Step 5: Create Issues

If user approves:

bash
# Detect repo for issue creation
REMOTE=$(git remote get-url origin 2>/dev/null)
# Extract owner/repo from remote URL
REPO=$(echo "$REMOTE" | sed -E 's|.*[:/]([^/]+/[^/]+?)(\.git)?$|\1|')

# For each actionable item:
gh issue create --repo "$REPO" --title "ITEM_TITLE" --body "From /forward handoff on YYYY-MM-DD"

Show results:

Created #115: Fix awaken git push auth
Created #116: /rrr --deep time-based

Step 6: Write to Outbox

Regardless of whether issues were created, write items to the outbox:

bash
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] || [ -L "$ORACLE_ROOT/ψ" ]; }; then
  PSI=$(readlink -f "$ORACLE_ROOT/ψ" 2>/dev/null || echo "$ORACLE_ROOT/ψ")
else
  PSI=$(readlink -f ψ 2>/dev/null || echo "ψ")
fi
OUTBOX_DIR="$PSI/outbox"
mkdir -p "$OUTBOX_DIR"

Write to: $PSI/outbox/YYYY-MM-DD_pending.md

markdown
# Pending Items — YYYY-MM-DD

## From: [repo-name] /forward

- [ ] Item 1 (issue #115)
- [ ] Item 2 (issue #116)
- [ ] Item 3 (no issue — skipped: vague)

Confirm outbox write (announce-mode — absolute paths required)

announce-mode → absolute path (no ψ/, no ~/, no $VAR, no ...).

Use: echo "marker: $RESOLVED_PATH" — bash substitutes. See CONVENTIONS.md.

bash
OUTBOX_FILE="$OUTBOX_DIR/$(date +%Y-%m-%d)_pending.md"
echo "📋 Outbox: $OUTBOX_FILE"

Silent Failures

  • If gh is not available: write to outbox only, skip issue creation silently
  • If repo has no GitHub remote: skip issue creation silently, write to outbox only
  • If gh auth status fails: skip issue creation silently, write to outbox only

Then: MUST Show Plan Approval Box

CRITICAL — DO NOT SKIP: The whole point of /forward is the plan approval UI. You MUST do ALL 3 steps in order. If you skip any step, the user cannot approve and clear the session.

  1. EnterPlanMode — enters plan mode
  2. Write plan file — session summary + next steps
  3. ExitPlanMode — THIS shows the approval box where user can approve/reject/clear

If you only do EnterPlanMode without ExitPlanMode, the user sees nothing. If you skip EnterPlanMode entirely, the user sees nothing. ALL 3 STEPS ARE REQUIRED.

Do NOT commit the handoff file — it lives in the vault, not the repo. After writing the handoff, gather cleanup context:

bash
# Check for things next session might need to clean up
git status --short
git branch --list | grep -v '^\* main$' | grep -v '^  main$'
gh pr list --state open --json number,title,headRefName --jq '.[] | "#\(.number) \(.title) (\(.headRefName))"' 2>/dev/null
gh issue list --state open --limit 5 --json number,title --jq '.[] | "#\(.number) \(.title)"' 2>/dev/null

Then:

  1. Call EnterPlanMode tool
  2. In plan mode, write a plan file with:
    • What we accomplished this session
    • Pending items carried forward
    • Cleanup needed (stale branches, open PRs, uncommitted files)
    • Next session goals and scope
    • Reference to handoff file path
    • Always end plan with a choice table:
markdown
## Next Session: Pick Your Path

| Option | Command | What It Does |
|--------|---------|--------------|
| **Continue** | `/recap` | Pick up where we left off |
| **Clean up first** | See cleanup list below, then `/recap` | Merge PRs, delete branches, close issues, then continue |
| **Fresh start** | `/recap --quick` | Minimal context, start something new |

### Cleanup Checklist (if any)
- [ ] [Open PR to merge]
- [ ] [Stale branch to delete]
- [ ] [Issue to close]
- [ ] [Uncommitted work to commit or stash]
  1. Call ExitPlanMode — user sees the built-in plan approval UI

The user gets the standard plan approval screen with options to approve, modify, or reject. This is the proper way to show plans.

If user calls /forward again — just show the existing plan, do not re-create the handoff file.

Wizard v2 Context in Handoff

If CLAUDE.md contains demographics from /awaken wizard v2, include in handoff:

markdown
## Context
**Oracle**: [name] ([pronouns]) | **Human**: [name] ([pronouns])
**Mode**: [Fast/Full Soul Sync] | **Memory**: [auto/manual]
**Team**: [solo/team context]

This helps the next session orient faster. If demographics not present, skip.


ASAP Mode

If user says /forward asap or /forward now:

  • Write handoff file
  • Immediately commit and push — no approval needed
  • Skip plan mode
  • User wants to close fast

Skip Plan Mode

If user says /forward --only:

  • Skip plan mode after commit
  • Just tell user: "💡 Run /plan to plan next session"

ARGUMENTS: $ARGUMENTS

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Forward AI skill do?

Hand off the current session to the next one. Use when user says "forward", "handoff", "wrap up", or before ending a session.

Why use Forward on TypingMind?

Because you install it once and use it with any model. Forward is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Forward in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Soul-Brews-Studio/arra-oracle-skills-cli/tree/alpha/skills/forward. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Forward?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Forward?

As many as you like. As long as a model supports skills, you can use Forward with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Forward AI skill free?

Yes. It is published on GitHub by Soul-Brews-Studio under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇