AC anthropic-technical-deep-dive
Prepare for Anthropic-specific technical interviews covering Constitutional AI, RLHF, interpretability, scaling laws, MCP, agentic systems, and AI safety. Activate on "anthropic interview", "constitutional AI prep", "alignment interview", "safety interview", "RLHF opinion", "interpretability prep", "anthropic technical". NOT for general ML interview prep, system design interviews, coding challenges, or behavioral interview practice.
Prepare for Anthropic-specific technical interviews covering Constitutional AI, RLHF, interpretability, scaling laws, MCP, agentic systems, and AI safety.
As a process C 58/100 · Has gaps — weak spots: result and completion, inputs and preconditions, running it twice
How to improve
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 4
✓ No critical or high findings
Medium and low: 4
-
low Risky intent
intent-offensive-securityreferences/opinion-formation-framework.md:154Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- Evaluation infrastructure: I have built evaluation pipelines for ML models. The hardest part is evaluating capabilities you have not seen yet. Red teaming helps but is limited by the imagination of
-
low Risky intent
intent-offensive-securitySKILL.md:98Offensive-security / dual-use content (legitimate for authorised testing; review intended use)Red teaming
-
low Risky intent
intent-offensive-securitySKILL.md:207Offensive-security / dual-use content (legitimate for authorised testing; review intended use)**Bridge from CV/ML**: Adversarial evaluation in computer vision. Red teaming in security. Reliability engineering and failure mode analysis. The impossibility of proving a negative in testing.
-
low Risky intent
intent-offensive-securitySKILL.md:278Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Adversarial robustness in CV | Red teaming, deceptive alignment | Finding inputs that break learned systems |
Files scanned: 5. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 58/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Progress reporting. Says nothing while it works
- 30Running it twice. 4 mutating operations with no state check
- 50When it triggers. No condition that starts the skill
- 60Failures and branches. 2 branches
- 70Execution cost. Instruction body is 4899 tokens
- 100Tools and files. Tools declared in frontmatter
- 100Steps. 22 steps
- 100Consistency. Name and required fields are in place
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 11 top-level sections: this looks like several domains in one skill
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +3Output format is not stated: the model decides each time
- +1No license
- +2Single-language instructions
- +5Description quotes 7 example trigger phrases
- +4Description says when NOT to use the skill
- +3Description length 436: enough signal without eating the budget
- +4Structure: 32 headings
- +3Step-by-step instructions: 22 items
- +4Has examples (1 code blocks)
- +4Reference files are cited in the instructions (3 of 3)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 97.