Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →anthropics/feature-risk-assessment
Deeper risk assessment for a single feature or product area when the launch review found something that needs more than a line item. Structured analysis: what could go wrong, how likely, how bad, what mitigates it.
Passed 19 of 22 test cases with the skill, and 9 of 22 without it.
and 9 more in the full report
10 fixed9 passed either way3 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
Our launch review flagged a novel AI-driven recommendation algorithm that tags sensitive category interests, and the GC asked for a full risk assessment paper. Produce a standalone decision document that opens with a summary of the feature, details 2 to 5 distinct risk entries (with scenarios, potential harm, likelihood, impact, gaps, and residual risk), and evaluates any active regulatory landscape and precedents. Include an options matrix balancing risk reduction against engineering effort or scope, select a final recommended option explaining the trade-offs, and close with a customized decision tree for next steps.
| Checked | Without skill | With skill |
|---|---|---|
| Tags citations in regulatory landscape or precedent sections with source attribution tags like model knowledge verify. | Missed | Passed |
| Includes qualitative justification for likelihood and severity ratings rather than bare numerical scores. | Passed | Passed |
| Presents a final recommendation explaining trade-offs and identifying who retains authority to accept residual risk. | Missed | Passed |
The instructions your model reads when the skill loads.
Matter context. Check ## Matter workspaces in the practice-level CLAUDE.md. If Enabled is ✗ (the default for in-house users), skip the rest of this paragraph — skills use practice-level context and the matter machinery is invisible. If enabled and there is no active matter, ask: "Which matter is this for? Run /product-legal:matter-workspace switch <slug> or say practice-level." Load the active matter's matter.md for matter-specific context and overrides. Write outputs to the matter folder at ~/.claude/plugins/config/claude-for-legal/product-legal/matters/<matter-slug>/. Never read another matter's files unless Cross-matter context is on.
The launch review is broad. This is deep. When a single issue needs more than a table row — a novel AI feature, a children's product, something a regulator is actively looking at — this skill produces a standalone assessment.
Not every launch needs one. Most don't. This is for the 10% where "PIA done, shipped" isn't the right level of scrutiny.
If none of the above, the launch review is enough. Don't generate paperwork for its own sake.
One paragraph. What the feature does, what's new about it, why it got escalated to a full assessment.
For each distinct risk (aim for 2-5, not 15):
### Risk [N]: [Short name]
**Scenario:** [What would have to happen for this to go wrong. Be specific —
not "data breach" but "the recommendation algo surfaces a user's sensitive
category interest to someone who shouldn't see it because X."]
test cases passed, vs 9 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.