Name: agent-memory
Rating: 64 (1 reviews)
Author: alirezarezvani

agent-memory

Use when a project's CLAUDE.md has grown past what anyone reads and you want the agent to learn durable facts from its own sessions instead — or when asking why the agent keeps re-learning the same correction, why a remembered rule is wrong, or where a memory line came from. Implements a four-tier store (L0 transcripts / L1 candidates / L2 project context / L3 stable persona) where promotion is earned by recurrence across sessions and days, never by one confident statement, and nothing reaches a committed file without a human adopting it.

Quality

78%

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 ./.gemini/skills/agent-memory/SKILL.md

The canonical home for this skill is agent-memory in alirezarezvani/claude-skills

Quality

Content

77%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 is well-organized, dense, and unusually strong on workflow clarity with thorough validation for a destructive-adjacent operation. Its weakness is bundle completeness: every external reference (two scripts and DESIGN.md) points to files that are not shipped, which both breaks progressive-disclosure navigation and leaves the otherwise-concrete commands non-executable.

Suggestions

Bundle the invoked scripts: create scripts/memory_inspect.py and scripts/memory_promote.py (or add a note that they are generated at install time) so the copy-paste commands in 'Use it' actually run.

Ship the referenced ../../DESIGN.md (rationale, open decisions, field schema) or relocate that detail into a bundled reference file so the one-level-deep link resolves instead of dangling.

After restoring the files, re-verify each code block and the DESIGN.md link resolves from the skill directory, which would lift both actionability and progressive_disclosure.

Dimension	Reasoning	Score
Conciseness	The body is efficient — a compact tier table, numeric gate thresholds ("≥ 3 distinct sessions spanning ≥ 2 distinct calendar days"), and lean code blocks assume Claude's competence — with only minor discursive framing in "The problem" and "Forcing questions" that could be trimmed, fitting anchor 4 rather than 'every token earns its place' at 5.	4 / 5
Actionability	Concrete, parameterized commands with example arguments are given ("python3 scripts/memory_inspect.py --why \"PR base branch is dev\"", "--contested", "python3 scripts/memory_promote.py"), but they invoke scripts that are not present in the bundle, so they are not fully runnable as-shipped — a minor executability gap that caps at anchor 4 rather than 'fully executable' at 5.	4 / 5
Workflow Clarity	The L0→L1→L2→L3 promotion sequence is clearly gated, with explicit validation ("Redact before writing", staged-only proposals, adopt-with-backup, dry-run "writes nothing", freeze-on-open-contradiction, fail-open hooks) and feedback loops ("--contested" → human resolves → promote); the five "Hard rules" act as a checklist, and because destructive-op validation is present the batch/destructive cap at 3 does not apply — anchor 5.	5 / 5
Progressive Disclosure	In-body structure is good (clear sections) and the delegation to [../../DESIGN.md](../../DESIGN.md) is cleanly signaled one level deep, but the actual bundle is empty — none of scripts/memory_inspect.py, scripts/memory_promote.py, or ../../DESIGN.md exist — so the progressive-disclosure navigation is broken and a reader cannot reach the delegated detail, pulling it below anchor 4.	3 / 5
	Total	16 / 20 Passed

Description

78%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 pairs a concrete 'what' (four-tier recurrence-gated store) with multiple natural 'when' triggers. Its main weakness is a second-person voice slip that costs it a specificity point; distinctiveness is good but not absolute given acknowledged sibling skills.

Dimension	Reasoning	Score
Specificity	Lists several concrete mechanisms ("four-tier store (L0 transcripts / L1 candidates / L2 project context / L3 stable persona)", "promotion is earned by recurrence across sessions and days", "nothing reaches a committed file without a human adopting it"), which is anchor-4 coverage, but the second-person phrasing "you want the agent to learn durable facts" triggers the mandated −1 voice penalty, dropping it to 3.	3 / 5
Completeness	An explicit "Use when…" clause plus three concrete "or when asking why…" triggers answers 'when', and "Implements a four-tier store … promotion is earned by recurrence … nothing reaches a committed file without a human adopting it" answers 'what' concretely — both present with concrete trigger phrases, matching anchor 5.	5 / 5
Trigger Term Quality	Natural trigger phrases a user would actually say are present ("CLAUDE.md has grown past what anyone reads", "why the agent keeps re-learning the same correction", "why a remembered rule is wrong", "where a memory line came from"), giving good keyword coverage with only a few obvious synonyms absent — anchor 4, not 5 because coverage is not comprehensive with synonyms/extensions.	4 / 5
Distinctiveness Conflict Risk	The four-tier recurrence-gated niche is distinct and the triggers are specific to memory auditing, but the skill's own frontmatter names sibling skills (memory-engineering, llm-wiki, skillopt-sleep), indicating minor overlap risk with closely related memory skills — anchor 4 rather than a clean niche at 5.	4 / 5
	Total	16 / 20 Passed

Validation

81%

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 — 13 / 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: 1 suspicious	Warning
referenced_paths_exist	Referenced path issues: 4 missing	Warning

	Total	13 / 16 Passed

Repository: alirezarezvani/claude-skills
Path: .gemini/skills/agent-memory/SKILL.md
Commit: 19392f7

Reviewed: ~2 hours 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.