Skill Self Evolution Enhancer
ReviewAudited by ClawScan on May 10, 2026.
Overview
The skill is transparent about adding self-evolution, but it can permanently change other skills and persistent agent instructions, so it needs careful review before use.
Install only if you want persistent self-evolution behavior added to other skills. Before using it, back up the target skill, run the generator with --dry-run, inspect all file changes, avoid overwriting existing EVOLUTION.md or .learnings files unintentionally, and do not add SOUL.md/AGENTS.md snippets unless you want the behavior to apply workspace-wide.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Running the script on the wrong skill, or on a skill that already has evolution files, could replace existing notes or behavior rules.
The optional generator writes and overwrites files in the chosen target skill directory without checking whether existing learning logs or EVOLUTION.md already exist, and without prompting for confirmation or creating backups.
cp "$ASSETS_DIR/LEARNINGS-TEMPLATE.md" "$LEARNINGS_DIR/LEARNINGS.md" cp "$ASSETS_DIR/ERRORS-TEMPLATE.md" "$LEARNINGS_DIR/ERRORS.md" cp "$ASSETS_DIR/FEATURE_REQUESTS-TEMPLATE.md" "$LEARNINGS_DIR/FEATURE_REQUESTS.md" ... cp "$ASSETS_DIR/EVOLUTION-RULES-TEMPLATE.md" "$TARGET_PATH/EVOLUTION.md"
Use --dry-run first, back up the target skill, inspect a diff before applying changes, and require explicit confirmation before overwriting any existing files.
Incorrect, sensitive, or maliciously shaped learning entries could be reused later and change how a skill responds.
The generated skills are instructed to persist prior learnings and reuse them in future tasks. The artifacts do not define validation, redaction, trust boundaries, or user approval before stored entries influence later behavior.
Before task: Load relevant entries from `.learnings/LEARNINGS.md` (and ERRORS.md if applicable). Filter by area, tags, or keywords. During task: Apply learnings when relevant.
Treat .learnings entries as untrusted until reviewed, redact private details, require approval before promotion to rules, and keep learnings scoped to the specific target skill.
A change intended for one skill could influence broader agent behavior across the workspace or future sessions.
SOUL.md and AGENTS.md are workspace-level agent instruction files. Adding self-evolution behavior there can affect future tasks beyond the single target skill.
Add these snippets to SOUL.md or AGENTS.md in the OpenClaw workspace (or to the enhanced skill's EVOLUTION.md) so OpenClaw proactively surfaces evolution feedback.
Prefer keeping evolution instructions inside the target skill's EVOLUTION.md unless the user explicitly opts into workspace-wide behavior changes.
Users may not realize there is a runnable helper script in addition to the instructions.
The package has no declared source/homepage and is described as instruction-only, but it includes a runnable shell helper. The script content is visible and purpose-aligned, so this is a provenance/review note rather than evidence of malicious behavior.
Source: unknown Homepage: none ... No install spec — this is an instruction-only skill. ... 1 code file(s): scripts/generate-evolution.sh
Review the included script before running it and install only from a source you trust.