CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

interrogate

Use for "interrogate", "adversarial review", "multi-model review", "challenge this", "stress test this code", "find blind spots", or "tear this apart". Adversaries try to break a change or design, and only findings proven by a run survive.

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

92%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

An exceptionally lean orchestration skill: a coherent four-phase workflow with an explicit run-based verification gate, a copy-paste adversary brief, and precise substitution rules for reusing the arena skill. The only real weakness is that its executability leans on the arena skill's phase text and an adapter doc that resolves outside the skill bundle.

Suggestions

Inline the minimal mechanics of the reused arena steps (how to spawn each adversary and what Phase A steps 3-4 actually do) so the core fan-out loop is executable without opening another skill, keeping the full arena text as the external reference.

The adapter link (../../docs/harness/claude-code.md) resolves outside the skill bundle and is not present in this copy; inline the essential Claude Code mappings or ship them under references/ so the skill is self-contained.

DimensionReasoningScore

Conciseness

About 25 lean lines where every sentence is instruction — "Skip arena Phases C, D, and E" and the one-line substitution rules assume Claude's competence with zero padding or explanation of known concepts. Not 4 because there is no over-explanation left to trim; every token earns its place.

5 / 5

Actionability

Provides a verbatim, copy-paste-ready adversary brief and specific execution details ("Spawn each adversary as `generalPurpose`", "Read the `interrogate reviewers` line in place of `arena runners`"). Not 5 because the core fan-out and verify steps are indirect pointers into the arena skill's phases rather than self-contained commands, so execution depends on holding another skill's text; not 3 because the guidance given is concrete and specific, well above pseudocode level.

4 / 5

Workflow Clarity

Clear Frame → Fan out → Verify → Output sequence with an explicit validation checkpoint: "Verify each finding per arena Phase F by rerunning its command. Keep a finding when the output shows the failure and the failure breaks the intent. Drop the rest." The batch fan-out of multiple adversaries includes the verify-and-drop feedback loop, so the batch-operation cap does not apply. Not 4 because validation is explicit with keep/drop criteria rather than implicit.

5 / 5

Progressive Disclosure

Under 50 lines with no bundle files needed and four clearly labeled sections (Frame, Fan out, Verify, Output); detail is appropriately deferred one level deep to clearly signaled targets (the arena skill, the Claude Code adapter doc), so the simple-skill exception applies. Not 4 because nothing that belongs in a separate file is inlined and every deferral is explicitly named.

5 / 5

Total

19

/

20

Passed

Description

86%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: explicit, trigger-phrase-rich 'when' guidance paired with a clear, third-person statement of what the skill does. The main gap is that the what-clause characterizes the approach rather than enumerating the skill's concrete actions, and a couple of triggers overlap a generic code-review skill.

Suggestions

Enumerate the skill's concrete actions in the what-clause — e.g. "Spawns multiple adversarial reviewers in isolated worktrees, verifies each finding by rerunning its command, and reports only run-proven failures" — to lift specificity beyond 1-2 characterizations.

Narrow the triggers that overlap a general code-review skill: qualify "stress test this code" and "challenge this" with "adversarially" or "to break it", keeping the distinctive phrases like "interrogate" primary.

DimensionReasoningScore

Specificity

Names the domain (adversarial review of a change or design) with two concrete behaviors — "Adversaries try to break a change or design" and "only findings proven by a run survive" — but does not enumerate the skill's actual actions (spawning adversaries, rerunning commands, reporting verified findings). Not 4 because there is no list of several specific actions; not 2 because the stated actions are concrete rather than generic.

3 / 5

Completeness

Explicitly answers 'when' with a concrete "Use for ..." clause listing trigger phrases, and 'what' with a clear statement of behavior and the survival rule for findings. Not 4 because the 'when' clause is more explicit and trigger-phrase-rich than the anchor-4 pattern of a generic 'Use when working with X'; both halves are clearly and concretely present.

5 / 5

Trigger Term Quality

Seven varied natural trigger phrases span the synonym space, from colloquial ("tear this apart", "challenge this", "stress test this code") to formal ("adversarial review", "multi-model review", "interrogate"). Not 4 because coverage already includes both casual and formal variants a user would naturally say; no file extensions are relevant to this process skill.

5 / 5

Distinctiveness Conflict Risk

Has a clear adversarial niche with distinctive triggers ("interrogate", "adversarial review", "tear this apart"), but "stress test this code" and "challenge this" could also plausibly trigger a general code-review skill. Not 5 due to this minor overlap with a closely related skill; not 3 because the run-proven-findings framing keeps the niche well defined.

4 / 5

Total

17

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

relative_links

Relative link issues: 1 suspicious

Warning

Total

15

/

16

Passed

Repository

cursor/plugins

Path

harness/pstack/claude-code/skills/interrogate/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.