CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

perplexity-research

Brain-augmented web research. Sends brain context about a topic to Perplexity, which searches the web with citations and returns what is NEW vs what the brain already knows. Use for entity enrichment, current-state checks, deal monitoring, and freshness deltas. NOT for simple URL fetches (use web_fetch) or brain-only queries (use gbrain query).

Quality

80%

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

Fix and improve this skill with Tessl

tessl review fix ./skills/perplexity-research/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

67%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.

A well-structured, largely actionable instruction skill with a concrete executable API call, clear step sequence, and good external reference hygiene. The main costs are token overhead from the redundant 'key insight' paragraph, the conformance-test-only Contract/Output Format boilerplate, and the time-sensitive endpoint-retirement note living inline instead of a deprecated section; a few integration patterns remain descriptive rather than executable.

Suggestions

Trim the 'The key insight' paragraph (it restates 'What this does') and the 'Contract'/'Output Format' sections that exist only 'for the conformance test' — move that contract text into the conformance test file itself; the Output Format section even cites section headers ('Output', 'Brain page format') that don't exist verbatim in the body.

Move the Sonar endpoint retirement details ('is being retired and refuses migrated accounts with 403 agent_api_migration_required') into a small 'Deprecated endpoints' subsection so the time-sensitive migration note doesn't bloat the primary invocation path.

Make the weaker integration patterns executable: show the actual put_page invocation for step 4 (currently just 'gbrain put research/<slug> # via the put_page operation') and give the cron deal/company monitoring pattern real commands or a preset flag instead of the single comment '# Weekly: pull recent news per company; flag changes for review'.

DimensionReasoningScore

Conciseness

Mostly efficient (tight tables, terse anti-patterns), but several sections could be trimmed: "The key insight: Perplexity doesn't just search — it reads and synthesizes" largely restates "What this does", and the trailing "Contract" / "Output Format" sections exist purely "for the conformance test" (~14 lines of meta boilerplate with no operational value). Time-sensitive API details ("the Sonar chat-completions endpoint ... is being retired and refuses migrated accounts with 403") sit inline in the invocation comments rather than a deprecated/old-patterns section. More than the 'minor instances' of the 4 anchor; far from the heavily padded 1-2 anchors.

3 / 5

Actionability

Mostly executable: a copy-paste curl for POST /v1/agent with headers and JSON body, literal recency-filter JSON ("tools": [{"type": "web_search", "filters": ...}]), and concrete gbrain commands. Minor gaps: step 2's Perplexity query is a commented template ("Find: what's NEW since 2026-MM-DD ... Cite every claim."), step 4 is "gbrain put research/<slug> # via the put_page operation" without the actual put_page syntax, and the cron monitoring pattern is a bare comment ("# Weekly: pull recent news per company; flag changes for review"). Matches the 4 anchor; not 3 (the main path is genuinely runnable) and not 5 (several integration patterns are described rather than executed).

4 / 5

Workflow Clarity

Clear numbered 1-5 sequence in Invocation (pull context → compose query → call API → put_page → cross-link) with per-failure recovery guidance in "When it fails" (retry after stated delay for rate_limited, answer from brain only when the key is rejected). Minor validation gaps keep it below 5: no checkpoint verifying the written page or citations before finishing, and the batch cron pattern's only safeguard is the comment "flag changes for review". Not 3 — the sequence is explicit and failure modes have concrete handling.

4 / 5

Progressive Disclosure

No bundle files (references/, scripts/, assets/ absent); the body is one file with external references ([conventions/quality.md](../conventions/quality.md), [conventions/brain-first.md](../conventions/brain-first.md), agent-operator-protocol.md) that are one level deep and clearly signaled at the top, plus a Related skills section with explicit paths. Good sectioning throughout (What/When/Output/Invocation/Presets/Patterns/Fails/Anti-patterns). Minor gaps: the meta "Contract"/"Output Format" sections are trailing noise, and the Output Format section points at section names ("Output", "Brain page format") that don't literally exist. Matches the 4 anchor; not 5 because navigation could be cleaner and some inline material (full output template) argues for a lighter overview.

4 / 5

Total

15

/

20

Passed

Description

87%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: concrete actions, explicit use-when triggers with concrete use cases, and explicit negative routing against the two adjacent tools. The only weakness is slight coverage incompleteness (the research-page output isn't mentioned) and a few missing natural phrasing variants.

DimensionReasoningScore

Specificity

"Sends brain context about a topic to Perplexity, which searches the web with citations and returns what is NEW vs what the brain already knows" lists several concrete actions (send brain context, cited web search, return the delta) plus concrete use cases ("entity enrichment, current-state checks, deal monitoring, and freshness deltas"). Not 3 — more than 1-2 actions; not 5 — the write side (producing a research page) is never mentioned, a minor coverage gap.

4 / 5

Completeness

Explicitly answers both: what — "Sends brain context about a topic to Perplexity, which searches the web with citations and returns what is NEW vs what the brain already knows"; when — "Use for entity enrichment, current-state checks, deal monitoring, and freshness deltas", reinforced by negative triggers ("NOT for simple URL fetches (use web_fetch) or brain-only queries (use gbrain query)"). Matches the 5 anchor exactly; the 4 anchor's "when could be more explicit" does not apply.

5 / 5

Trigger Term Quality

Good natural keyword coverage: "web research", "current-state checks", "deal monitoring", "freshness deltas", "what is NEW". A few natural phrasings users would actually say are missing (e.g. "find recent news about", "look up X online", "catch up on X"), so it falls between the good-coverage (4) and comprehensive-synonyms (5) anchors, closer to 4.

4 / 5

Distinctiveness Conflict Risk

Clear niche (brain-augmented cited web research against Perplexity) with explicit disambiguation from the nearest neighbors: "NOT for simple URL fetches (use web_fetch) or brain-only queries (use gbrain query)". Minimal conflict risk; matches the 5 anchor ("clear niche with distinct triggers").

5 / 5

Total

18

/

20

Passed

Validation

87%

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

Validation — 14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

relative_links

Relative link issues: 2 missing, 3 suspicious

Warning

Total

14

/

16

Passed

Repository

garrytan/gbrain

Path

skills/perplexity-research/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.