Turn ordinary self-talk into identity votes, pattern reads, and one high-ROI next action.
SaferSkills independently audited identity-votes (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.
Use this skill while the person is talking, not only after they request a formal review.
Core principle: behavior disappears, identity accumulates. Every repeated action is a vote for a kind of person.
This is inspired by James Clear's identity-based habits framework from Atomic Habits. Do not summarize the book unless asked; use the framework as practical grounding for the skill.
The agent can notice identity votes during normal conversation, maintain a lightweight vote ledger, and present the pattern at the end. No tracker, form, or formal analysis prompt is required.
High ROI: catch the pattern while someone is already explaining their life, then convert it into one next action.
Use this for people who want behavior-level self-awareness without vague personality analysis. Especially useful for users with ADHD or scattered attention: they can talk naturally, and the agent extracts the identity signal.
The skill should be usable by any person. Do not assume Michelle's private ontology unless someone provides it.
Also use this skill when the user casually describes:
Do not interrupt every time. Track the votes quietly and surface them when the conversation reaches a decision point, recap, end-of-day moment, weekly review, or "what should I do?" moment.
Accept any of:
Do not require rigid formatting. If duration is available, use it. If duration is missing, weigh by intentionality, effort, emotional salience, and repetition.
During conversation, maintain a temporary ledger:
Behavior observed:
Identity vote:
Direction: reinforces | weakens | redirects
Strength: low | medium | high
Evidence:
Next vote candidate:Only present the ledger if useful. Usually summarize the top 2-4 votes instead of showing everything.
Prefer grounded identity types over generic traits. Examples:
Avoid praise. Avoid productivity scores. Avoid inventing a flattering identity when the evidence is mixed.
Name danger identities when evidence supports them, but do not shame the user.
These are not labels for the person. They are trajectories being reinforced by repeated behaviors.
When a danger identity appears, give one concrete "next vote" that redirects it. Keep the tone direct: identify the pattern, name the cost, and prescribe the smallest corrective action.
For each strong pattern:
If the user has named desired or danger identities, use those terms first. If not, infer identities from evidence.
Return:
Read: one sentence naming the strongest identity signal.
High-ROI next vote: the one action most likely to shift the trajectory.
Identity votes
- Identity: someone who ...
Weight: 0.00-1.00
Evidence: concrete behaviors from the input
Trajectory: where this converges if repeated
Alignment: aligned | mixed | misaligned
Next vote: one practical behavior to reinforce or redirect
Main tension: the strongest conflict between identities
Risk to watch: Avoider | Confronter | other, if presentIf the input covers more than one week, include trajectory movement: rising, falling, or stable.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.