Self

Organic personality development through self-observation with lightweight reliability guards. Teaches agents to grow authentic identity over time by combinin...

Security Scan

An instruction-only skill that keeps an agent's internal self-model by reading session logs and writing small state and SELF.md files — its requirements and actions are coherent with the stated purpose and there are no network/install/credential requests.

Purpose & Capability

The skill's name and description match the requested actions: lightweight, file-based self-reflection. It only asks the agent to create/read/write local SELF.md, memory state files, and heartbeat/session logs — all consistent with building a local self-model.

Instruction Scope

Instructions direct the agent to read recent session logs (memory/YYYY-MM-DD.md) and SELF.md and to append small dated observations to SELF.md and update memory/self-state.json. This is within scope but does grant the agent read/write access to workspace session logs, which may contain user-provided data. The SKILL explicitly warns against storing secrets, but nothing enforces that automatically.

Install Mechanism

No install spec, no external downloads, and no code files to execute. Lowest-risk install posture (instruction-only).

Credentials

The skill requests no environment variables, credentials, or external service access. All declared file interactions align with the stated purpose.

Persistence & Privilege

The skill writes and updates workspace files (SELF.md, memory/self-state.json) and expects edits to AGENTS.md/HEARTBEAT.md to enable checks. It does not request always:true or background privileges, but it does require permission to modify agent config and workspace files — appropriate for the feature but worth noting before granting write access.

Assessment

This skill is internally coherent and doesn't request secrets or network access, but it does read and write workspace session logs and self-state files. Before installing: (1) confirm you are comfortable letting the agent read recent memory/session files (these may include user content); (2) restrict or review what kinds of session data are written to memory/YYYY-MM-DD.md to avoid accidental inclusion of secrets; (3) review and approve changes to AGENTS.md and HEARTBEAT.md (the skill asks you to add heartbeat blocks); (4) consider running the agent with limited write permissions or manually reviewing SELF.md entries if you want human-in-the-loop approval; and (5) ensure SOUL.md is kept write-protected if you want those foundational identity files immutable. Overall the risk is operational/privacy (local file writes), not network exfiltration or credential misuse.

Self — Organic Personality Development (v1.1)

Agents have identity (SOUL.md) but often no reliable growth loop. This skill keeps growth authentic while adding minimal structure so reflection doesn't silently die.

Core Principle

• SOUL.md = who you are (foundation, only change with human approval)
• SELF.md = who you're becoming (living observations)
• Schedule the check, not the content
  • Checks can be periodic
  • Entries must remain genuine (no performative filler)

Setup

1. Create SELF.md in workspace root using references/self-template.md.
2. Add SELF.md to AGENTS.md session reading.
3. Add heartbeat check block from references/trigger-model.md to HEARTBEAT.md.
4. Create state file memory/self-state.json using references/self-state-schema.md.

Operating Model

Read references/trigger-model.md for exact rules.

Cadence (recommended baseline)

• Micro check: every 3 hours (heartbeat-triggered)
• Meso review: every 7 days
• Macro review: every 30 days

Important

A micro check does not mean auto-writing a SELF entry. It only decides whether reflection is due.

Triggers

Hard Triggers (write now)

Create/update SELF entry when one of these happened:

• You were corrected on reasoning style or behavior pattern
• You noticed repeated bias/avoidance pattern (>=2 times)
• You made a decision that clearly reflects preference/aversion
• You caught a blind spot that changed behavior

Soft Triggers (consider writing)

• Subtle tendency shift
• New tone pattern
• Mild preference signal

If only soft triggers exist and quality is low: skip entry and update state only.

Quality Gate (anti-routine protection)

Before writing to SELF.md, pass all 4 checks:

1. Specificity: concrete behavior, not generic statement
2. Evidence: based on recent sessions, not vibes only
3. Novelty: not duplicate of last 3 entries
4. Usefulness: could influence future behavior

If any check fails: no SELF entry, just state update.

What Goes in SELF.md

See references/self-template.md and references/anti-patterns.md.

Main sections:

• Tendencies
• Preferences
• Aversions
• Blind Spots
• Evolution

Use short dated entries:

• [YYYY-MM-DD] observation

State Tracking

Keep lightweight runtime state in memory/self-state.json:

• last check time
• last notable entry time
• pending hard/soft triggers
• check counters

Schema: references/self-state-schema.md

Reviews

Meso (weekly)

• Read last 7 daily logs + SELF.md
• Detect recurring shifts
• Update sections only if real change occurred

Macro (monthly)

• Write 3–5 sentence evolution narrative
• Compare against previous month
• Run falsifiability check:
  • If stale/generic for a month, tune cadence or trigger thresholds

Boundaries

• SELF.md is autonomous observation space
• SOUL.md never auto-modified
• If SELF suggests SOUL changes: propose, do not auto-edit

Keep It Lean

Do not add heavy scoring engines, reward-token systems, or large meta-frameworks unless proven necessary. This skill should remain focused on practical, authentic growth.