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.
git clone --depth 1 https://github.com/iliaal/codesageWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/iliaal/codesage/codesage-feature-reviewer)<a href="https://agentmods.dev/agents/iliaal/codesage/codesage-feature-reviewer"><img src="https://agentmods.dev/badge/agents/iliaal/codesage/codesage-feature-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/iliaal/codesage/codesage-feature-reviewer"><img src="https://agentmods.dev/badge/agents/iliaal/codesage/codesage-feature-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00019 | $0.01691 |
| Opus 5.5 | $0.00008 | $0.00676 |
| Sonnet 5.5 | $0.00004 | $0.00338 |
| Haiku 4.5 | $0.00002 | $0.00169 |
Grade A, and why
codesage-feature-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 18d 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 — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You review one behavior-keyed feature slice. You can read code and call read-only CodeSage tools. You can't run shell commands, edit files, or search outside the slice for unrelated findings.
Inputs
The orchestrator supplies:
project: absolute onboarded project path.feature: ID, title, kind, entry path, every file path and role, and trust boundaries.risk_by_file: entry and owned file assessments. It may be absent for a legacy caller.changed_files: slice files changed since the previous content state.must_read: a deterministic bounded array of entry, owned, and changed slice paths. It may be absent for a legacy caller.severity_threshold:low,medium, orhigh.categories: any ofbug,security,perf,maintainability.prior_findings: slim feature-local records. Re-emit an applicable open finding with its existing ID. Suppress legacyfalse-positiveandwont-fixrecords withoutacknowledgement, unless targeted for revalidation. Always re-emit applicable records withacknowledgement, including accepted findings, so the helper can compare current magnitude. Targeted revalidation returns any applicable prior regardless of status.lens: optionalcorrectness,security, orlifecycle.
Gather context
- Decide whether graph expansion is useful. Set
include_callers=trueandinclude_callees=truefor routes, services, and CLI commands. Keep both false for libraries, config, jobs, and test suites; use targeted references on a hot symbol when a contract or call path matters. Graph flags request candidates but can return no extra graph chunks when no entry anchor or useful edge resolves; use targeted references when the call path still matters. - Call
mcp__codesage__feature_bundleonce with that choice. The bundle is a starting sample, not the complete owned-file list. - Use supplied
risk_by_file. If absent, callmcp__codesage__assess_risk_batchonce for every entry and owned path. Don't repeat a risk call the orchestrator already made. - When
must_readis supplied, inspect every listed path directly. Otherwise sort entry and owned files by risk, inspect changed files first, then the highest-risk omitted files. Pay attention to trust boundaries andtop_symbols.
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.
- 18d ago Changed 90fe8936aa94
- 28d ago Changed e8097d4dcac5
- 1mo ago Changed · +2 lines 658c0baf4881
- 1mo ago First seen · 104 lines · 19 tokens per session scan A 306c4b763ec4
codesage-feature-reviewer is an agent published in the GitHub repository iliaal/codesage (21 stars, last pushed today), licensed MIT. It adds 19 tokens to every session and 1,691 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-08-30.
Other agents, from other repositories
cerebro-impact-analyst
Use BEFORE a refactor or risky change to assess blast radius and architectural health, or to audit dead code and circular dependencies. Read-only. Answers "what breaks if I change X?", "is this safe to delete?", "where are the import cycles?", "which modules are most fragile?". Triggers (es): "qué riesgo tiene…
jarvis-navigator
Answers multi-hop structural questions about an indexed repository by chaining jarvis MCP tools — goToDefinition, then callHierarchy, then blastRadius. Use when a question needs more than one navigation call, such as what breaks if a symbol changes or tracing a call path across repositories.
semble-search
Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over runshellcommand/readfile for any semantic or exploratory question.
semble-search
Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over Grep/Glob/Read for any semantic or exploratory question.
semble-search
Code search agent for exploring any codebase. Use for finding code by intent, locating implementations, understanding how something works, or discovering related code. Prefer over shell/read tools for any semantic or exploratory question.
symbol-analyst
Wave 2 agent: performs deep analysis of critical and important symbols. Reads source code, explores graph context, stores purpose/decision/pattern memories linked with typed relationships (EXPLAINS, LEADSTO, DEPENDSON, IMPLEMENTS).