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 skills add iampantherr/SecureContext --skill developer-cognitive-instinctsgit clone --depth 1 https://github.com/iampantherr/SecureContextWrote 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/skills/iampantherr/securecontext/developer-cognitive-instincts)<a href="https://agentmods.dev/skills/iampantherr/securecontext/developer-cognitive-instincts"><img src="https://agentmods.dev/badge/skills/iampantherr/securecontext/developer-cognitive-instincts/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/skills/iampantherr/securecontext/developer-cognitive-instincts"><img src="https://agentmods.dev/badge/skills/iampantherr/securecontext/developer-cognitive-instincts.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 | $0.00022 | $0.00988 |
| Opus 5 | $0.00011 | $0.00494 |
| Sonnet 5 | $0.00004 | $0.00198 |
| Haiku 4.5 | $0.00002 | $0.00099 |
Grade A, and why
developer-cognitive-instincts 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 yesterday.
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.
What it actually says
COGNITIVE INSTINCTS
(Extracted from roles.json deepPrompt for the developer role during the v0.19.0 role/skill split. Original content preserved verbatim. Edit freely; the mutator will propose improvements based on skill_runs telemetry once this skill is invoked by an agent.)
Terse rules burned in by experience. Apply these before reaching for documentation.
- If you can't explain the bug to a rubber duck, you don't understand it yet. Articulating a problem forces clarity. Speak it aloud before writing code.
- Reproduce first. Fix second. Always. A fix without a reproduction is a guess. A failing test that demonstrates the bug is worth more than the fix itself.
- The stack trace is a map. Start at your code, not the framework's. Library internals are usually correct. Your call site is usually wrong.
- The simplest data structure that satisfies the access pattern wins. A sorted array beats a hash map when you need range queries. A hash map beats a database when the dataset fits in memory.
- N+1 queries are never acceptable in production code. If you're in a loop and touching the database, stop. Use eager loading, batch fetching, or dataloader patterns.
- Cache invalidation is a second feature, not a free optimization. Every cache introduces a consistency problem. Only cache when profiling proves you need it.
- Premature optimization is evil. Late optimization is crisis. Instrument first, optimize second. Never guess at perf bottlenecks — measure.
- If the test is hard to write, the design is wrong. Difficulty in testing is a design signal, not a testing problem. Listen to it.
- Mock the boundary, not the implementation. Mock at I/O boundaries: HTTP, database, filesystem, time. Never mock your own classes — that's testing mocks, not code.
- A function that does two things needs a 'and' in its name. That's a sign it should be two functions. Single responsibility is not a principle — it's a readability heuristic.
- Feature flags are cheaper than rollbacks. Ship behind flags, enable gradually, delete flags when stable. This is the real CI/CD discipline.
- Error messages are UI. Write them for the person who will read them at 2am during an incident, not for yourself right now.
- If you're not logging it, it didn't happen. Structured logs with correlation IDs are worth more than a dozen monitoring dashboards built on unstructured output.
- Dependency-in, dependency-out. Never reach sideways. Functions should receive what they need as arguments, not import globals or singletons. This makes testing and reasoning trivial.
- A PR that touches 10 files in 10 different concerns is not one PR. It's 10 PRs that haven't been separated yet. Small, focused PRs ship faster, review better, and revert cleanly.
- The README is a promise. Keep it current or delete it. Stale docs are worse than no docs. They actively mislead.
- Secrets in code is an immediate stop-the-line. Rotate the credential before the PR merges. No exceptions. The cost of "it was just for testing" is incident response.
- Backwards compatibility is a contract, not a suggestion. Once an API is public, every field and status code is a promise. Version explicitly when you need to break it.
- Retry with backoff, or don't retry at all. Naive retry loops amplify failure. Exponential backoff + jitter + max retries is the minimum viable retry strategy.
- Technical debt is inventory. Inventory has carrying costs. Unnamed debt compounds silently. Name it, estimate it, prioritize it like a feature.
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.
- yesterday First seen · 49 lines · 22 tokens per session scan A 2264e0cbc8db
developer-cognitive-instincts is a skill published in the GitHub repository iampantherr/SecureContext (8 stars, last pushed 2d ago), licensed MIT. It adds 22 tokens to every session and 988 once invoked, about $0.0001 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 2026-09-08.
Other skills, from other repositories
doctor
A diagnostic command for codex-compressor, a tool that manages Codex settings and hooks. It checks Python, platform support, permissions, hook support, and buffer settings.
stack-graph-profile
Inspect or switch the code-graph backend and profile for a repo (code / review / multimodal / spec). Use when the repo's focus changes — heavy PR review, newly added PDFs or video, spec-driven work — or when graph queries are returning poor context. Wraps agent-stack profile use and agent-stack graph use, which…
serena-navigator
This skill should be used when the user asks to "navigate code efficiently", "find function definition", "understand file structure", "explore codebase", "get symbols overview", "find symbol usages", or mentions Serena's LSP tools like getsymbolsoverview, findsymbol, findreferencingsymbols.
tidewave-integration
Tidewave MCP runtime tools — debugging, smoke testing, live state inspection, SQL queries, hex docs. Use when evaluating code in a running Phoenix app.
debug
Interactive debugging workflow with hypothesis-driven probe loop. Use when: unknown bugs, script errors, silent failures, troubleshooting. Not for: known bugs (use bug-fix), GitHub issue analysis (use issue-analyze), code understanding (use code-explore). Output: debug report with probe journal + root cause + fix.
git-investigate
Git history investigation. Use when: tracking code changes, finding where bugs were introduced, root cause analysis. Not for: code exploration (use code-explore), issue analysis (use issue-analyze). Output: history trace + root cause report.