correction-pipeline

When the user corrects a factual error, root-cause it immediately. Don't just note the correction — trace the error to its source, fix the source, and prevent recurrence. Every factual error is either a data error (bad brain page, bad memory file, bad rendered SOUL/USER identity, bad facts row) or a hallucination (LLM confabulated from partial signals).

Quality

73%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./plugin-variants/gbrain-coding/skills/correction-pipeline/SKILL.md

Quality

Content

75%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 highly actionable, well-sequenced workflow with concrete commands and real validation checkpoints for its destructive operations. The chief weakness is conciseness — Anti-Patterns and Contract sections restate guidance already embedded in the steps — and the absence of any bundle files means a long single-file body with no offloading reference.

Suggestions

Collapse the Anti-Patterns section: each anti-pattern already maps to a Step 6 rule, so replace the prose list with a one-line 'see Step 6' pointer or merge them as a negative column in the Step 5 classification table.

Move the Severity Tiers and Recurring Error Patterns tables into a references/ file (e.g. patterns.md) referenced once from the body, shrinking the always-loaded SKILL.md toward an overview.

Add an explicit 'if the read-back does not match the user's words, redo the --set step before --confirm' feedback loop to the SOUL_USER_ERROR fix so error recovery is stated, not implied.

Dimension	Reasoning	Score
Conciseness	The body is dense and domain-specific (no generic concepts Claude already knows), but several sections are noticeably redundant — the Anti-Patterns block restates Step 6 rules (don't edit SOUL.md directly, don't fix only memory, re-sync after repo edits) and the Contract restates the step guarantees, which could be trimmed.	3 / 5
Actionability	Fully executable, copy-paste-ready commands with flags and provenance throughout ('gbrain search', 'grep -ri', 'gbrain recall --grep', 'gbrain forget <fact-id>', 'gbrain remember ... --provenance', 'gbrain bootstrap interview --set/--confirm', 'gbrain bootstrap render --force'), covering the common error classes concretely.	5 / 5
Workflow Clarity	Eight steps are explicitly ordered with a classification decision gate (Step 5) and verification checkpoints (Step 7 propagation grep, the bootstrap read-back/--confirm hash ritual, Step 8 report); because validation is present the destructive cap does not apply, though error-recovery feedback loops are mostly implicit rather than spelled out.	4 / 5
Progressive Disclosure	No bundle files ship with the skill, but the body is well-sectioned and delegates detail to sibling skills via clearly signaled one-level references ('The full interview discipline ... lives in skills/soul-audit/SKILL.md — route through it'); minor gap is that long ancillary tables (Severity Tiers, Recurring Patterns) stay inline in a ~250-line file with no reference file to offload them.	4 / 5
	Total	16 / 20 Passed

Description

71%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 specific, well-scoped description that clearly states what it does and when it applies, with a strong distinct niche. Its main weakness is that natural user-facing trigger phrases are absent from the description text (relegated to the frontmatter triggers list), leaving the 'when' less concrete than it could be.

Suggestions

Fold one or two concrete trigger phrases into the description itself (e.g. 'Use when the user says "that's wrong" or "I never said that" and corrects a fact') so the 'when' carries natural keywords, not just the separate triggers list.

Tighten the action verbs from process language ('root-cause it', 'trace the error') toward the concrete fix surfaces already named, so the 'what' reads as a discrete action list rather than a workflow summary.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete actions ('root-cause it immediately', 'trace the error to its source, fix the source, and prevent recurrence') plus a concrete taxonomy of error surfaces ('bad brain page, bad memory file, bad rendered SOUL/USER identity, bad facts row' or 'hallucination'), but the verbs are somewhat high-level compared to the granular action lists in the 5 anchor.	4 / 5
Completeness	Both 'what' (root-cause, trace, fix source, prevent recurrence) and 'when' ('When the user corrects a factual error') are present and clear, but the 'when' lacks concrete trigger phrases in the description text, so it stops short of the explicit-trigger anchor 5.	4 / 5
Trigger Term Quality	The description carries relevant domain keywords ('factual error', 'corrects', 'hallucination', 'confabulated') but omits the natural phrases a user would actually say ('that's wrong', 'that's not true') — those live in the separate frontmatter triggers field, not in the description itself.	3 / 5
Distinctiveness Conflict Risk	It carves a clear niche — root-causing user-corrected factual errors in a brain/knowledge system with a specific data-vs-hallucination taxonomy — making it unlikely to fire for an unrelated skill.	5 / 5
	Total	16 / 20 Passed

Validation

87%

Warnings & errors only

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

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 2 suspicious	Warning

	Total	14 / 16 Passed

Repository: garrytan/gbrain
Path: plugin-variants/gbrain-coding/skills/correction-pipeline/SKILL.md
Commit: 43597b1

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

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.