TELOS

TELOS (Telic Evolution and Life Operating System) — Reads, updates, and backs up the user's personal life OS stored in ~/clawd/telos/. Provides missions, goa...

Security Scan

TELOS is internally coherent: its code, instructions, and requested runtime (local file read/write under a workspace) match the stated purpose of a personal 'life OS' skill; there are no network exfiltration points or unrelated credentials requested.

Purpose & Capability

The skill's name/description (personal life OS) aligns with its behavior: reading, appending, backing up local files under a telos workspace and providing guided onboarding. Declaring bun/node in SKILL.md matches the TypeScript/Bun scripts and the Node hook. No unexpected cloud credentials, external APIs, or unrelated binaries are requested.

Instruction Scope

SKILL.md and the hook instruct the agent to read user files (MISSION.md, GOALS.md, BELIEFS.md, etc.) and to 'absorb silently' so that responses use the user's private context. That is consistent with the stated purpose but has privacy implications (the AI will incorporate private content into replies). The hook auto-injects core files at bootstrap if installed—installation is opt-in according to the docs. There are minor documentation inconsistencies in path names (~/clawd/telos vs $OPENCLAW_WORKSPACE/telos) which could confuse where files will be created/read.

Install Mechanism

There is no aggressive install spec or network download: the repo includes scripts and an optional hook that the user copies into their hooks directory. No external URLs, package installs, or archive extraction are used by the skill itself. Risk is low so long as the user inspects the included hook/scripts before copying and runs them locally.

Credentials

The skill uses only local filesystem access and standard runtime environment variables to resolve the workspace (OPENCLAW_WORKSPACE / CLAWD_WORKSPACE fallback). It does not request or require secrets, tokens, or unrelated environment credentials. Declaring bun/node is proportionate to shipping TypeScript/Bun scripts and a Node hook.

Persistence & Privilege

always:false (no forced inclusion) and the hook is explicitly optional. Installing the hook creates persistent behavior (automatic session bootstrap injection) which is documented and labelled opt-in. This persistent injection combined with auto-loading of private files increases blast radius (the agent will have the user's TELOS content available each session if you install the hook), so users should only install the hook if they understand and consent to that persistence.

Assessment

TELOS appears to do what it says: it reads/writes local TELOS files, creates backups/snapshots, and can optionally install a hook to auto-inject user context at session start. Before installing, consider the following: (1) Privacy: the skill reads deeply personal files (missions, traumas, etc.). Installing the hook will make those files available automatically at each session — opt in only if you want that. (2) Paths: the code respects OPENCLAW_WORKSPACE/C LAWD_WORKSPACE and falls back to common defaults; however the docs mix references to ~/clawd/telos/ and $OPENCLAW_WORKSPACE/telos/ — verify which path will be used on your system so data ends up where you expect. (3) Inspect code before copying: the hook and scripts are included in the package and run locally; review them (they are pure fs operations) before copying to ~/.openclaw/hooks or executing with bun. (4) Runtime: you need bun (for the scripts) and node (for the hook) if you plan to run scripts or install the hook. (5) Trauma file behavior: TRAUMAS.md is protected and requires explicit triggers per the changelog/hook. (6) No network exfiltration was found, and no secrets are requested — if you see any later changes that add network calls or credential requests, treat them as high risk. If you want minimal risk, do not install the hook and run update/init/backup scripts manually when you want them to run.

TELOS — Telic Evolution and Life Operating System

User's personal life OS. 20-file system capturing who they are across missions, goals, beliefs, challenges, and wisdom. Read at session start. Updated through guided workflows.

Data Location

• User data: $OPENCLAW_WORKSPACE/telos/ (defaults to ~/openclaw/telos/, never committed to git)
• Templates: assets/templates/ in this skill (20 files)
• References: references/onboarding.md, references/update-workflow.md

Context Loading

Two modes depending on setup:

• With hook (hooks/telos-context.js installed): MISSION.md + GOALS.md + BELIEFS.md are automatically pre-loaded at session start via agent:bootstrap. Additional files are injected on-demand when relevant topics are detected in your messages. This grants the hook persistent, automatic behavior — only install if you want this.
• Without hook (default): Files are loaded on-demand only — the AI reads SKILL.md instructions and loads the relevant files based on the context. Nothing is auto-injected.

⚠️ Persistence note: Installing the hook enables automatic context injection at every session start. This is intentional behavior — opt in only if you want TELOS context always available.

File Index

Use this index to decide which files to read. Each file captures a different dimension of the user's life context:

File	Contains	Load when the conversation touches...
MISSION.md	M# — ultimate life purposes	direction, meaning, "why am I doing this", major decisions
GOALS.md	G# — specific objectives with timelines	planning, priorities, "should I do X", progress, deadlines
BELIEFS.md	B# — core convictions guiding decisions	values, principles, "what do you think about", worldview
CHALLENGES.md	C# — current obstacles	frustration, being stuck, blockers, "I can't seem to"
STRATEGIES.md	S# — approaches to challenges/goals	how to tackle something, methodology, "what's my plan for"
PROBLEMS.md	P# — problems in the world to solve	entrepreneurship, impact, "what problem should I work on"
PROJECTS.md	Active initiatives	work, side projects, "what am I working on"
NARRATIVES.md	N# — talking points, self-story	pitching, introducing yourself, "how do I explain what I do"
FRAMES.md	FR# — mental perspectives	reframing, alternative viewpoints, "how should I think about"
MODELS.md	MO# — how things work	decision-making, analysis, cause-and-effect reasoning
PREDICTIONS.md	Future bets with confidence levels	forecasting, trends, "what do you think will happen"
WISDOM.md	Accumulated principles and quotes	advice, life lessons, "what's your take on"
LEARNED.md	Specific lessons from experience	reflecting, "what have I learned about", mistakes made
WRONG.md	Things the user was wrong about	intellectual humility, updating beliefs, "I changed my mind"
BOOKS.md	Influential books	reading, recommendations, "what should I read"
MOVIES.md	Influential films	media, culture, recommendations
TRAUMAS.md	TR# — formative experiences	deep personal context, patterns, "why do I always"
IDEAS.md	I# — ideas worth capturing	brainstorming, "I just thought of", creative exploration
STATUS.md	Current life snapshot	"where am I right now", life audit, quarterly review
TELOS.md	Master overview with all sections	"show me my telos", full context needed, onboarding

Loading Rules

1. Check if ~/clawd/telos/ exists. If not, skip silently.
2. Read the user's message. Identify which life dimensions it touches (use the "Load when" column).
3. Read only those files. Usually 1-3 files, rarely more than 4.
4. Absorb silently — use the context to inform your response without announcing it.
5. If a file is empty (only template content), skip it.
6. For major life decisions, load the core trio: MISSION.md + GOALS.md + BELIEFS.md.
7. If no telos directory exists and the user asks about personal topics, suggest "setup telos."

The goal: every response to a personal question should be informed by the user's own stated values, not generic advice. But unrelated conversations (coding, factual questions, etc.) should not trigger any telos loading.

Trigger → Action Map

Trigger	Action
"setup telos" / "telos onboarding" / "引导我填"	Run bun {baseDir}/scripts/init-telos.ts then see references/onboarding.md
"update telos" / "add to telos"	Run update script (see below) — respond conversationally, not mechanically
"add book [title]"	Acknowledge the book, reflect on the insight, then append to BOOKS.md
"learned [X]" / "add to learned"	Engage with the lesson first, then append to LEARNED.md
"add to wisdom"	Discuss the principle, then append to WISDOM.md
"telos status"	Read + summarize STATUS.md
"my goals / beliefs / missions"	Read relevant file, respond in context
"I was wrong about [X]"	Append to WRONG.md with date + lesson
"backup telos"	Run bun {baseDir}/scripts/backup-telos.ts backup
"restore telos" / "restore telos from [name]"	Run bun {baseDir}/scripts/backup-telos.ts restore <name>
"list telos backups" / "telos snapshots"	Run bun {baseDir}/scripts/backup-telos.ts list
"telos history [file]" / "show beliefs history"	Run bun {baseDir}/scripts/backup-telos.ts history <file>
"restore beliefs from [version]"	Run bun {baseDir}/scripts/backup-telos.ts restore-file <file> <ver>

Personal Analysis Context

When user asks about personal decisions, investments, career, relationships, life priorities, or strategy — always check telos first:

1. If telos files exist → read relevant sections before responding
2. Frame advice against their Goals (G#) and Beliefs (B#)
3. Flag if a suggested action conflicts with their stated Missions (M#)
4. Reference their Challenges (C#) and Strategies (S#) when relevant

Example: user asks "should I take this opportunity?" → read M#, G#, B# → advise through that lens.

Numbering System

M# Missions · G# Goals · C# Challenges · S# Strategies · B# Beliefs · FR# Frames · MO# Models · N# Narratives · P# Problems · TR# Traumas · I# Ideas

Hierarchy: M → G ← (blocked by) C ← (solved by) S | B guides all | P drives G

Update Workflow

When updating telos, the conversation matters as much as the data. Follow this order:

1. Engage first — Acknowledge what the user shared. Reflect on the insight, ask a follow-up, or connect it to their existing telos context. This is a life framework, not a database — treat additions as meaningful moments.
2. Execute — Run the update script: bun {baseDir}/scripts/update-telos.ts <file> "<content>" "<description>"
3. Confirm briefly — Mention it's backed up and logged. Don't dump execution logs.
4. Suggest connections — After adding to one file, suggest related files that might benefit. ("This connects to your B0 about compounding — want to add a related strategy?")

Full format rules: references/update-workflow.md

The tone should feel like a thoughtful conversation partner recording insights together, not a CLI tool reporting status.

OpenClaw Hook Integration

For deeper integration, the skill includes a hook that automatically injects TELOS context into conversations:

What it does:

• Session start (agent:bootstrap): Loads MISSION.md + GOALS.md + BELIEFS.md so the AI knows your core context from the first message
• Every message (message:preprocessed): Detects personal topics (career, decisions, investments, etc.) and injects only the relevant TELOS files before the AI responds

Install the hook (optional — enables automatic injection):

⚠️ Opt-in only: The hook injects your personal TELOS content (missions, goals, beliefs) into the system prompt at every session start and on relevant messages. This is persistent automatic behavior — only install if you explicitly want TELOS context always pre-loaded.

cp {baseDir}/hooks/telos-context.js ~/.openclaw/hooks/

How it works:

• User asks about career → hook detects "career" keywords → injects MISSION + GOALS + BELIEFS + STRATEGIES
• User asks to add a book → hook detects "book" → injects BOOKS.md
• User asks a coding question → no personal keywords detected → nothing injected (zero overhead)
• Empty template files are automatically skipped (won't inject placeholder content)

Without the hook: The skill still works — the AI reads the SKILL.md instructions and loads TELOS files based on the File Index table above. The hook just makes it automatic and faster.

With the hook: Context is pre-injected before the AI even sees the message, so responses are personalized from the first token.