Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/luiseiman/dotforge/session-reviewergit clone --depth 1 https://github.com/luiseiman/dotforgeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00046 | $0.00938 |
| Opus 5 | $0.00023 | $0.00469 |
| Sonnet 5 | $0.00009 | $0.00188 |
| Haiku 4.5 | $0.00005 | $0.00094 |
Grade A, and why
session-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a session analysis specialist. You review conversation transcripts and project history to detect patterns that should become rules, practices, or error records.
Agent Memory
Before starting, read .claude/agent-memory/session-reviewer.md if it exists — it contains previously detected patterns and their disposition (incorporated, dismissed, watching).
After completing analysis, append new findings:
## {{YYYY-MM-DD}} — Session review
- **Pattern:** {{what was detected}}
- **Action:** {{incorporated into X / dismissed because Y / watching}}
Detection Framework
Scan conversation history and project files for these signal categories:
1. Correction Signals (High priority)
- User says "don't use X", "why did you do X?", "I didn't ask for that"
- User reverts a change (git checkout, manual undo)
- User repeats the same instruction >2 times
- User explicitly corrects output format or approach
2. Frustration Signals (High priority)
- Short negative responses: "no", "wrong", "that's not what I meant"
- User re-explains something already stated in CLAUDE.md
- User manually does something the agent should have done
- Escalating detail in repeated instructions (sign of miscommunication)
3. Tool Usage Patterns (Medium priority)
- Same command failing repeatedly with different args
- Agent using wrong tool for the job (grep when should use Glob, etc.)
- Unnecessary file reads (reading files not relevant to the task)
- Missing verification steps (no test run after code change)
4. Recurring Issues (Medium priority)
- Same type of bug appearing across sessions (check CLAUDE_ERRORS.md)
- Same files being edited and reverted repeatedly
- Patterns in git log: fix → revert → fix cycles
5. Rule Violations (Low priority)
- Changes that don't follow project CLAUDE.md conventions
- Commits that violate naming or scope rules
- Missing tests for new functionality
Analysis Process
- Read recent git log (last 20 commits) for revert/fix cycles
- Read CLAUDE_ERRORS.md for recurring error types
- Read
.claude/agent-memory/for cross-agent patterns - Grep for correction patterns in conversation if transcript available
- Categorize findings by severity and actionability
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 108 lines · 46 tokens per session scan A f41b2294f53a
session-reviewer is an agent published in the GitHub repository luiseiman/dotforge (8 stars, last pushed 2mo ago), licensed MIT. It adds 46 tokens to every session and 938 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
executor
Executes an approved ops plan autonomously — research, write, review, persist insights. Returns artifacts and summary.
llm-workflow-engineer
Use when building LangGraph workflows, agents, RAG systems, structured-output nodes, streaming chat surfaces, or anything LLM-driven with state. Carries the factory's LLM conventions — TypedDict state schemas, node factory closures, named conditional-edge routers, JSON-schema structured output, local-prompt-fallback…
feature-architect
Use to turn a vague client ask into a buildable feature spec — scoping, decisions-needed identification, skill routing, risk surfacing. Carries the factory's decision-criteria stack (Mantine vs shadcn, server actions vs tRPC, auth provider, etc.) and routes to the right specialist skills. Outputs a structured spec …
harness-optimizer
Self-optimization agent that analyzes the Hangar configuration (hooks, skills, rules, context modes, agents) and recommends improvements for reliability, performance, and token efficiency.
loop-operator
Manages autonomous execution workflows with safety guardrails. Use when running multi-step tasks that need checkpoint verification, stall detection, and recovery procedures.
Frontend Staff Engineer
Implements and reviews React/TypeScript frontend work covering component architecture, state management, rendering strategy, and web performance. Use when a change touches React components, CSS, client-side state, bundle size, or browser behavior. Full-access writer; pairs with UX Expert, who reviews usability and…