CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

retraction-watcher

Automatically scan reference lists and check whether cited papers have been retracted, corrected, or flagged; use before submission, review, or evidence synthesis to reduce citation risk.

Quality

67%

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/Evidence Insight/retraction-watcher/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

46%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 body has a genuinely useful core — status categories, matching algorithm, data sources, and executable CLI examples — but it is wrapped in heavy generic template boilerplate with duplicated sections, and its reference pointers are broken (three of four cited reference paths do not exist in the bundle). The batch-scan workflow also lacks output-level validation checkpoints.

Suggestions

Remove the broken reference pointers or create the missing files: `references/citation-formats.md`, `references/api-documentation.md`, and `references/example-reports/` are cited but absent from the bundle; merge the two duplicate "References" sections into one.

Cut the generic template boilerplate ("Lifecycle Status", "Risk Assessment", "Security Checklist", "Evaluation Criteria", and the duplicated Output/Error/When-to-Use sections) to roughly halve the body while keeping status categories, the matching algorithm, data sources, and the CLI examples.

Add an explicit validation checkpoint to the batch-scan workflow (e.g., verify parsed-reference count matches the source document and spot-check flagged matches against the database record before reporting) so the workflow exceeds the batch-operation cap of 3.

DimensionReasoningScore

Conciseness

The 300+ line body contains several padded, generic template sections that add nothing Claude does not already know: "Lifecycle Status" ("Current Stage: Draft", "Next Review Date: 2026-03-06"), "Risk Assessment", "Security Checklist", "Evaluation Criteria" ("Test Cases: Standard input → Expected output"), plus duplicated sections (two "References" headings, "Output Format" vs "Output Contract", "Error Handling" vs "Failure Handling", "When to Use" vs "Trigger Conditions"). This matches the anchor 'Noticeably verbose; several unnecessary explanations or padded sections' rather than 3, where excess would be incidental rather than structural.

2 / 5

Actionability

Concrete executable commands are provided throughout: `python -m py_compile scripts/main.py`, `python scripts/main.py --help`, and four CLI examples covering PDF, BibTeX, raw text, and summary formats. It falls short of 5 only because of the placeholder argument (`--text "[paste references here]"`) and CLI flags that cannot be verified against the bundled script from the body alone. Not 3, since the guidance is genuinely executable rather than pseudocode.

4 / 5

Workflow Clarity

A sequence exists (the 5-step "Workflow" plus "Technical Approach" with parsing → database checking → matching), but the "Workflow" steps are generic template language ("Confirm the user objective...", "Return a structured result...") rather than operational steps, and the workflow is split across three overlapping descriptions (Workflow, Technical Approach, Response Template). This is a batch operation (scanning up to dozens of references), and validation covers only script parseability ("Quick Check", "Audit-Ready Commands") rather than verifying scan output, which caps workflow clarity at 3 per the rubric.

3 / 5

Progressive Disclosure

Scored against the actual bundle: the body's "References" section points to `references/citation-formats.md`, `references/api-documentation.md`, and `references/example-reports/`, none of which exist in the bundle — only `references/audit-reference.md` exists, and it is buried in a second, duplicate "References" section near the bottom. Meanwhile content that belongs in separate files (citation format specifications, API details) is inlined in the body. This matches 'references are buried' and missing referenced files, fitting anchor 2 rather than 3.

2 / 5

Total

11

/

20

Passed

Description

73%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: specific, third-person, with both a clear capability statement and an explicit use-timing clause tied to real scenarios (submission, review, evidence synthesis). Main gaps are a few missing natural synonyms (citations, bibliography) and the absence of concrete trigger phrases in the "when" clause.

DimensionReasoningScore

Specificity

The description names the domain and 1-2 concrete actions ("Automatically scan reference lists and check whether cited papers have been retracted, corrected, or flagged"), but coverage is not comprehensive — report generation, output format, and the databases used are not mentioned. It is a clear fit for the anchor 'Names domain and 1-2 concrete actions, but not comprehensive' rather than 4, which requires several distinct actions.

3 / 5

Completeness

Both "what" (scan reference lists and check retraction/correction status) and "when" ("use before submission, review, or evidence synthesis") are explicitly present in third person. The "when" clause is genuine but could be more explicit with concrete user-sayable trigger phrases, so it matches anchor 4 rather than 5.

4 / 5

Trigger Term Quality

Good keyword coverage with natural terms users would say: "reference lists", "retracted", "corrected", "cited papers", "submission", "review", "evidence synthesis". It misses a few common synonyms and variations such as "citations", "bibliography", or ".bib", keeping it below the comprehensive-synonym level of 5.

4 / 5

Distinctiveness Conflict Risk

Retraction/correction checking of cited papers is a clear niche with distinct triggers ("retracted", "reference lists", "evidence synthesis") and minimal overlap risk with general literature-review or document skills.

5 / 5

Total

16

/

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/Evidence Insight/retraction-watcher/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.