learn-from-session — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited learn-from-session (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Examine what was accomplished in this session, focusing on:
permissions.allow in ~/.claude/settings.json)Beyond rules updates, check for opportunities across all Claude Code configuration:
| Category | What to look for |
|---|---|
| CLAUDE.md | Universal conventions, workflow preferences, common pitfalls, anti-patterns ("tried X, doesn't work because Y, do Z instead") |
| Memory topic files | Domain-specific knowledge → route to the right topic file (check memory/ dir for existing files; keep MEMORY.md for universal rules and pointers only) |
| Slash Commands | Repetitive multi-step tasks (3+ commands), common workflows |
| Skills | Domain-specific guidance, code review rules, conventions |
| Agents | Specialized workflows, multi-step analysis tasks |
| Hooks | Pre-commit checks, build verification, auto-formatting |
| Permissions | Commands that required manual approval — check against existing allowlist in ~/.claude/settings.json under permissions.allow |
Present results in this format. Each finding must be concrete — include the specific command, file, or pattern observed.
## Session Summary
**What we did:**
- [Brief bullet points of main accomplishments]
**Patterns observed:**
- [Recurring actions, questions, or friction points]
---
## Extracted Learnings
- Ran `arc lint -a` 4 times after editing — add a PostToolUse hook for auto-formatting
- `buck test` failed twice because target was `test_foo` not `foo_test` — add naming convention to CLAUDE.md
- Permission prompt for `jf submit` appeared 3 times — add to allowlist
---
## Improvement Opportunities
### Rules Updates
- [specific rule text to add or change, with target file path]
### Anti-Patterns Discovered
- **[What was tried]** → [Why it failed] → [Do this instead]
- Target file: [which memory/CLAUDE.md file to add this to]
### Memory Updates
- [specific entry to add/update] → Target: [topic file name]
### Slash Commands
- **`/[name]`** — [use case], Priority: [High/Medium/Low]
### Skills
- **`[name]`** — [domain], Priority: [High/Medium/Low]
### Hooks
- **[name]** — [type], Priority: [High/Medium/Low]
### Permissions
- **`Bash(command:pattern*)`** — [what it does], approved N times (in allowlist: yes/no)After presenting findings, ask:
"Would you like me to apply any of these changes now? I can: - Update memory files (MEMORY.md or topic files) - Add permissions to settings.json - Create/update CLAUDE.md entries"
Apply only the changes the user approves. Do NOT auto-apply without confirmation. For complex CLAUDE.md rule edits, use /llm-rules-updater instead.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.