CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

image-ocr

Extract text from images with Tesseract OCR; use it when you need to recognize text from PNG/JPEG/TIFF/BMP images, select a language model, or run OCR via natural-language requests (e.g., "Interpret the image at C:\path\image.png").

Quality

71%

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 ./scientific-skills/Other/image-ocr/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

42%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.

The core sections (When to Use, Example Usage, Implementation Details) are concrete and the structure is orderly, but the body is padded with nine sections of skill-agnostic boilerplate, documents a config path and a --help behavior that the bundled script does not implement, and never links the troubleshooting reference it ships. It reads like a filled-in template rather than documentation of the actual bundle.

Suggestions

Fix the config path mismatch: image_ocr.py loads artifacts/ocr_config.json (CONFIG_PATH), so change every mention of scripts/ocr_config.json in the body to match, or change the script — otherwise the documented edit-then-run workflow silently no-ops.

Delete the generic template sections (When Not to Use, Required Inputs, Recommended Workflow, Deterministic Output Rules, Output Contract, Validation and Safety Rules, Failure Handling, Completion Checklist, Scope Reminder) — none contain OCR-specific content — and replace the fabricated --help "Quick Validation" with the script's real success/failure outputs (e.g. exit codes 2-5 and their messages).

Link references/ocr-troubleshooting.md from the body (e.g. a "Troubleshooting" pointer in Implementation Details) and remove the duplicated inline guidance about tesseract_cmd/PATH and language packs so the reference file is discoverable instead of orphaned.

DimensionReasoningScore

Conciseness

Roughly half the body is generic template boilerplate with no OCR-specific content — "When Not to Use", "Required Inputs", "Recommended Workflow", "Deterministic Output Rules", "Output Contract", "Validation and Safety Rules", "Failure Handling", "Completion Checklist", and a "Scope Reminder" that just repeats the description. The config keys are also documented twice (Example Usage configs plus Implementation Details), and "Dependencies" pads with "(version not specified)". This matches the "noticeably verbose; several unnecessary explanations or padded sections" anchor rather than the mostly-efficient 3.

2 / 5

Actionability

There is concrete, executable-looking guidance (pip install, two JSON config examples, run command), but it contains factual errors that break execution: the body says to edit `scripts/ocr_config.json` while the script actually loads `artifacts/ocr_config.json` (CONFIG_PATH in image_ocr.py), so documented edits are silently ignored; and "Quick Validation" runs `python scripts/image_ocr.py --help`, which the argparse-free script does not support. Not 4 because these are not minor gaps — following the documented workflow verbatim fails; not 2 because the config schema and run command are genuinely concrete.

3 / 5

Workflow Clarity

The execution flow is clearly sequenced (load config → resolve path → open image → OCR → output) and validation steps are present in abundance, but the primary checkpoint is fabricated: the "Expected output format" for the quick validation ("Result file: image_ocr_result.md / Validation summary / Assumptions") bears no relation to what the script actually prints (the extracted text or an error message). The documented happy-path workflow also misdirects users to the wrong config file, so checkpoints don't verify the real flow.

3 / 5

Progressive Disclosure

The body is well-sectioned and the script paths it cites (scripts/image_ocr.py, scripts/ocr_config.json) exist in the bundle, but `references/ocr-troubleshooting.md` is never linked from the body at all — it is an orphaned, buried reference — while its content (tesseract not on PATH, missing language pack, empty OCR output) is partially re-inlined into the body's Implementation Details. Structure exists but the reference is not signaled and content that belongs in it is inline, matching the 3 anchor.

3 / 5

Total

11

/

20

Passed

Description

88%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, third-person, with an explicit "use it when" clause and a real example trigger phrase. Its only weaknesses are a few missing natural synonyms ("read text", "transcribe") and an example utterance that could route image-interpretation requests to non-OCR skills.

DimensionReasoningScore

Specificity

Multiple concrete actions are explicitly stated — "Extract text from images with Tesseract OCR", "recognize text from PNG/JPEG/TIFF/BMP images", "select a language model", and "run OCR via natural-language requests" — covering the skill's full capability surface (extraction, formats, language selection, request parsing). Not 4 because no meaningful capability gap remains for this domain.

5 / 5

Completeness

Explicitly answers both questions: the "what" ("Extract text from images with Tesseract OCR") and the "when" ("use it when you need to recognize text from PNG/JPEG/TIFF/BMP images, select a language model, or run OCR via natural-language requests") with a concrete trigger phrase. Matches the score-5 anchor directly; third-person voice is maintained.

5 / 5

Trigger Term Quality

Good natural keyword coverage: "OCR", "Tesseract", "extract text", "recognize text", the format names PNG/JPEG/TIFF/BMP, and a concrete example utterance ("Interpret the image at C:\path\image.png"). Not 5 because common synonyms users would say — "read text from an image", "transcribe", "image to text" — are missing.

4 / 5

Distinctiveness Conflict Risk

The Tesseract-OCR-on-images niche is clear and largely distinct, but the trigger phrasing "Interpret the image at ..." reads like a general vision/image-interpretation request, creating minor overlap risk with multimodal image-reading skills rather than OCR specifically. Not 5 due to this ambiguity; not 3 since the domain (local OCR via Tesseract) remains well-differentiated.

4 / 5

Total

18

/

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

frontmatter_unknown_keys

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

Warning

Total

15

/

16

Passed

Repository

aipoch/medical-research-skills

Path

scientific-skills/Other/image-ocr/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.