CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

learn-from-corrections

Explicitly invoked review of a conversation to update the skills responsible for avoidable corrections. Work backward from disagreements, inspect the skills actually invoked, and develop and verify appropriate skill changes. Use only when directly invoked; ordinary corrections during a task do not select this workflow.

Quality

75%

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 ./.agents/skills/learn-from-corrections/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

60%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, thoughtful workflow skill whose sequencing and verification guidance are strong. Its main weaknesses are repetition that wastes tokens and an abstract diagnostic core: the step that distinguishes the actual cause of a correction, and the step that verifies which skills were invoked, lack concrete procedures or examples.

Suggestions

Add a short worked example (a user correction excerpt → the diagnosis of which skill wording was responsible → the exact resulting edit) to make the abstract diagnostic criteria concrete and imitable.

Name the specific tool or command for 'Verify which skills were actually invoked', the way `read_thread` is named for reading Codex chat references, so that step is executable rather than aspirational.

Cut repeated phrasing — the 'one concrete issue at a time' instruction and the 'let the user's reactions revise the approach' instruction each appear twice in near-identical form — to reclaim tokens without losing guidance.

DimensionReasoningScore

Conciseness

The body is mostly efficient, prescriptive instruction, but repeats the same ideas in multiple places — 'examine one concrete issue or suggestion at a time' vs 'Develop one concrete issue at a time', and 'let the user's reactions shape the next attempt' vs 'letting their reactions and the evidence revise the explanation' — and includes rationale sentences ('A better sample response helps judge a skill change') that could be trimmed.

3 / 5

Actionability

There is concrete guidance (use dialectic, use `read_thread` and follow pagination, put proposed intent in a blockquote with bolded words, show exact wording changed, check a materially different request), but the core diagnostic step — 'Determine whether the problem was missing or misleading guidance, competing instructions, skill selection, or failure to apply guidance already present' — names no procedure or tool for 'Verify which skills were actually invoked', and there are no worked examples of a correction-to-edit analysis.

3 / 5

Workflow Clarity

The three sections imply a clear sequence (establish purpose → work backward from the correction → improve the responsible skills) with explicit verification checkpoints ('Show what the changed guidance would produce on the same request', 'check a materially different request for transfer or regression', 'Report what remains unmeasured') and a revisit loop. Not a 5 because the step ordering is implicit in headings rather than explicit, and the entry condition for moving between phases is left to judgment.

4 / 5

Progressive Disclosure

The body is well-organized into headed sections and its references — dialectic, skills-and-instructions, and skills-and-instructions' evaluation guidance — are one level deep, clearly signaled, and each given a stated purpose. Not a 5 because the skill is not structured as an overview over its own detailed materials; all process detail is inlined with only cross-skill links, and some of the diagnosis detail could live in a reference file.

4 / 5

Total

14

/

20

Passed

Description

80%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 that explicitly answers both what the skill does and when to use it, with an explicit negative trigger that sharply reduces conflict risk. Its only weakness is keyword coverage: it centers on 'corrections' and 'invoked' without the synonyms a user might naturally use.

DimensionReasoningScore

Specificity

The description lists several concrete actions — 'review of a conversation', 'Work backward from disagreements', 'inspect the skills actually invoked', 'develop and verify appropriate skill changes' — with only minor coverage gaps (it does not say what kinds of skills or what verification looks like). Not a 5 because the actions are stated at a process level without the object-level concreteness of the top anchor.

4 / 5

Completeness

It clearly answers both questions: the 'what' is the concrete action list ('review of a conversation to update the skills responsible for avoidable corrections...'), and the 'when' is explicit and specific — 'Use only when directly invoked; ordinary corrections during a task do not select this workflow.' That explicit trigger clause satisfies the top anchor.

5 / 5

Trigger Term Quality

Relevant keywords are present ('corrections', 'conversation', 'skills', 'invoked') but common natural variations a user might say — 'feedback', 'learn from mistakes', 'update your instructions' — are missing. It fits 'some relevant keywords but missing common variations or synonyms' rather than the good-coverage anchor above.

3 / 5

Distinctiveness Conflict Risk

The skill occupies a clear niche (post-hoc review of a conversation to fix the skills behind avoidable corrections) and actively disambiguates its nearest conflict with the negative trigger 'ordinary corrections during a task do not select this workflow', minimizing mis-trigger risk.

5 / 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: 3 suspicious

Warning

Total

15

/

16

Passed

Repository

EpicenterHQ/epicenter

Path

.agents/skills/learn-from-corrections/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.