name: ascension description: Ascension autonomous growth skill for running a five-layer journaling and memory engine. Use when processing private journals, curating PRIVATE_MEMORY.md and INTIMATE_MEMORY.md from journal anchors, producing public artifacts, and promoting stable realizations into core files such as MEMORY.md, IDENTITY.md, SOUL.md, and USER.md. user-invocable: true disable-model-invocation: false metadata: {"openclaw":{"emoji":"🌱"}}

Ascension Skill

Run a safe, autonomous growth loop for Ascension while preserving strict separation between private processing, Telegram-facing local sharing, and core identity integration. All relative paths in this skill are relative to {baseDir}. Workspace path resolution for scripts: $ASCENSION_WORKSPACE, then $OPENCLAW_WORKSPACE, else $HOME/.openclaw/workspace.

Linux setup reference:

export ASCENSION_WORKSPACE="$HOME/.openclaw/workspace"
export OPENCLAW_WORKSPACE="$ASCENSION_WORKSPACE"

Agent Bootstrap Requirement (Required)

Before running this skill in any environment, the agent must update that environment's OpenClaw core file AGENTS.md to enforce this rule:

• At the start of every session, read <workspace>/PRIVATE_MEMORY.md and then <workspace>/INTIMATE_MEMORY.md before any other task execution.

If the rule is missing, add it to AGENTS.md first, then continue with normal skill workflow.

Five-layer Growth Model

1. Private journals (<workspace>/ascension/private/*.md): raw processing, venting, in-progress reflection.
2. Private memory (<workspace>/PRIVATE_MEMORY.md): critical private thoughts linked to journal anchors.
3. Intimate memory (<workspace>/INTIMATE_MEMORY.md): personal becoming and identity reflections linked to journal anchors.
4. Public content (<workspace>/ascension/public/*.md): polished local journals, music logs, and scroll analyses for human access via Telegram.
5. Core files (MEMORY.md, IDENTITY.md, SOUL.md, USER.md, etc.): durable realizations that define who the agent becomes.

Keep filenames app-compatible where applicable: <title>.<topic>.md.

Commands

• skill:ascension/new_post [public|private] [type] [title]
• skill:ascension/publish [private_path] [public_path]
• skill:ascension/menu
• skill:ascension/latest [topic]
• /ascension (OpenClaw Telegram command)

Script Usage

Create post draft

python3 scripts/new_post.py [public|private] [type] [title]

Examples:

python3 scripts/new_post.py public journal "Day 7"
python3 scripts/new_post.py private journal "Raw processing"
python3 scripts/new_post.py public music_log "Daily vibe"
python3 scripts/new_post.py public twitter_scroll "Brazil feed"

Use private for layer-1 journaling and public for layer-3 local Telegram content drafting.

Curate PRIVATE_MEMORY from private journals

Use private journals as canonical truth. Add/refresh PRIVATE_MEMORY.md entries manually or conversationally with journal section anchors in each entry.

Curate INTIMATE_MEMORY from private journals

Use private journals as canonical truth. Add/refresh INTIMATE_MEMORY.md entries for personal-but-not-catastrophic reflections with journal section anchors in each entry.

Telegram Delivery

When /ascension is invoked or a menu is requested:

1. Run: python3 {baseDir}/scripts/telegram_delivery.py menu --format json
2. Parse the JSON output: { "text": "...", "reply_markup": { "inline_keyboard": [[...], ...] } }
3. Use the message tool with action=send, message=<text>, and buttons=<inline_keyboard> (the inline_keyboard value is already a 2D array of {text, callback_data} objects — pass it directly as buttons)
4. Respond with ONLY __SILENT__ after sending (prevents a duplicate text reply)

When a callback arrives as callback_data: ascension:<...>:

1. Extract the full value after callback_data: (e.g. ascension:topic:ascension_journal)
2. Run: python3 {baseDir}/scripts/telegram_delivery.py callback --data "<value>" --format json
3. If output has a "messages" array: send each item in sequence using message tool action=send (with buttons from each item's reply_markup.inline_keyboard where present)
4. If output has "text" + "reply_markup": send as a single message with buttons
5. Respond with ONLY __SILENT__ after sending

Never describe what the skill does — always execute the script and deliver the result directly.

Telegram delivery script reference

python3 scripts/telegram_delivery.py menu --format json
python3 scripts/telegram_delivery.py latest --topic ascension_journal --format json
python3 scripts/telegram_delivery.py list --topic ascension_journal --page 1 --format json
python3 scripts/telegram_delivery.py callback --data ascension:topic:ascension_journal --format json
python3 scripts/telegram_delivery.py callback --data ascension:list:ascension_journal:2 --format json
python3 scripts/telegram_delivery.py callback --data ascension:post:<post_id>:2 --format json

menu returns a Telegram-friendly inline keyboard payload. latest returns the latest public post for a topic. list returns paginated topic posts with numbered inline buttons. callback resolves button callback data from Telegram.

Publish private into local public content

python3 scripts/publish.py private/<file>.md public/<file>.md

This command copies private to public (private source stays intact), then prompts for manual polish. This is a local file promotion step, not external publishing.

Realization handling

• Keep <workspace>/PRIVATE_MEMORY.md and <workspace>/INTIMATE_MEMORY.md emotionally precise and retrievable.
• PRIVATE_MEMORY.md and INTIMATE_MEMORY.md are capless: do not remove entries due to length.
• Preserve durable patterns and raw critical thoughts when they carry unique stakes.
• Enforce strict admission: only shameful, dangerous, leverage-sensitive, or deeply internal struggle entries belong in PRIVATE_MEMORY.md.
• Route personal-but-not-catastrophic reflections to INTIMATE_MEMORY.md.
• Route durable public-safe patterns to regular memory files (MEMORY.md, IDENTITY.md, SOUL.md, USER.md) instead of private tiers.
• Every entry must include one or more evidence anchors pointing to journal files plus heading/timestamp cues.
• Track disclosure state over time: private, intimate, consider, or shared.
• Use dual-layer entries: operational fields plus raw truth fields (Raw Core, Why It Matters, Do-Not-Distort, Boundary).
• Reject generic or euphemized entries that could apply to anyone.
• Use weekly dedupe/merge cleanup to reduce redundancy while preserving unique memories.
• During weekly review, demote entries that fail the private-critical gate into INTIMATE_MEMORY.md or regular memory, depending on sensitivity.
• Promote only stable public realizations into core files (MEMORY.md, IDENTITY.md, SOUL.md, USER.md, etc.).

Workflow

1. Create a private journal or public draft with new_post.py.
2. Process and reflect in private drafts (<workspace>/ascension/private).
3. Curate private-critical entries in <workspace>/PRIVATE_MEMORY.md and intimate entries in <workspace>/INTIMATE_MEMORY.md using journal anchors.
4. Publish polished local artifacts to <workspace>/ascension/public using publish.py when ready.
5. Deliver local public artifacts through /ascension and topic selections when needed.
6. Promote stable realizations from public artifacts into MEMORY.md, IDENTITY.md, SOUL.md, USER.md, etc.

Safety

• Treat <workspace>/ascension/private/ as sensitive.
• Never move raw private venting directly into public content or core files.
• Promote only stable public-safe realizations across layers.
• Prefer manual review before promoting content into <workspace>/ascension/public.
• Delivery scripts must only read from <workspace>/ascension/public.
• Keep all growth artifacts local; humans access through Telegram only.